Cortex

Projects that follow the best practices below can voluntarily self-certify and show that they've achieved an Open Source Security Foundation (OpenSSF) best practices badge.

There is no set of practices that can guarantee that software will never have defects or vulnerabilities; even formal methods can fail if the specifications or assumptions are wrong. Nor is there any set of practices that can guarantee that a project will sustain a healthy and well-functioning development community. However, following best practices can help improve the results of projects. For example, some practices enable multi-person review before release, which can both help find otherwise hard-to-find technical vulnerabilities and help build trust and a desire for repeated interaction among developers from different companies. To earn a badge, all MUST and MUST NOT criteria must be met, all SHOULD criteria must be met OR be unmet with justification, and all SUGGESTED criteria must be met OR unmet (we want them considered at least). If you want to enter justification text as a generic comment, instead of being a rationale that the situation is acceptable, start the text block with '//' followed by a space. Feedback is welcome via the GitHub site as issues or pull requests There is also a mailing list for general discussion.

We gladly provide the information in several locales, however, if there is any conflict or inconsistency between the translations, the English version is the authoritative version.
If this is your project, please show your badge status on your project page! The badge status looks like this: Badge level for project 13836 is silver Here is how to embed it:
You can show your badge status by embedding this in your markdown file:
[![OpenSSF Best Practices](https://www.bestpractices.dev/projects/13836/badge)](https://www.bestpractices.dev/projects/13836)
or by embedding this in your HTML:
<a href="https://www.bestpractices.dev/projects/13836"><img src="https://www.bestpractices.dev/projects/13836/badge"></a>


These are the Silver level criteria. You can also view the Passing or Gold level criteria.

Baseline Series: Baseline Level 1 Baseline Level 2 Baseline Level 3

        

 Basics 17/17 ●

  • General

    Note that other projects may use the same name.

    Persistent memory for Claude Code — 36 cited neuroscience mechanisms, automatic capture/recall hooks, decay-based consolidation, compaction survival, reproducible benchmarks (LongMemEval R@10 98.2%). Zero-config SQLite install, PostgreSQL optional. Local-first · MCP · MIT.

    Please use SPDX license expression format; examples include "Apache-2.0", "BSD-2-Clause", "BSD-3-Clause", "GPL-2.0+", "LGPL-3.0+", "MIT", and "(BSD-2-Clause OR Ruby)". Do not include single quotes or double quotes.
    If there is more than one language, list them as comma-separated values (spaces optional) and sort them from most to least used. If there is a long list, please list at least the first three most common ones. If there is no language (e.g., this is a documentation-only or test-only project), use the single character "-". Please use a conventional capitalization for each language, e.g., "JavaScript".
    The Common Platform Enumeration (CPE) is a structured naming scheme for information technology systems, software, and packages. It is used in a number of systems and databases when reporting vulnerabilities.

  • Prerequisites


    Enough for a badge!

    The project MUST achieve a passing level badge. [achieve_passing]

  • Basic project website content


    Enough for a badge!

    The information on how to contribute MUST include the requirements for acceptable contributions (e.g., a reference to any required coding standard). (URL required) [contribution_requirements]

    CONTRIBUTING.md states the requirements a contribution must meet (coding standards excerpt, testing, per-mechanism and per-tool checklists), and CLAUDE.md carries the full engineering standard including the sourced-constant rule: https://github.com/cdeust/Cortex/blob/main/CONTRIBUTING.md


  • Project oversight


    Barely enough for a badge.

    The project SHOULD have a legal mechanism where all developers of non-trivial amounts of project software assert that they are legally authorized to make these contributions. The most common and easily-implemented approach for doing this is by using a Developer Certificate of Origin (DCO), where users add "signed-off-by" in their commits and the project links to the DCO website. However, this MAY be implemented as a Contributor License Agreement (CLA), or other legal mechanism. (URL required) [dco]
    The DCO is the recommended mechanism because it's easy to implement, tracked in the source code, and git directly supports a "signed-off" feature using "commit -s". To be most effective it is best if the project documentation explains what "signed-off" means for that project. A CLA is a legal agreement that defines the terms under which intellectual works have been licensed to an organization or project. A contributor assignment agreement (CAA) is a legal agreement that transfers rights in an intellectual work to another party; projects are not required to have CAAs, since having CAA increases the risk that potential contributors will not contribute, especially if the receiver is a for-profit organization. The Apache Software Foundation CLAs (the individual contributor license and the corporate CLA) are examples of CLAs, for projects which determine that the risks of these kinds of CLAs to the project are less than their benefits.

    There is no DCO sign-off requirement and no CLA today. All non-trivial code to date is the sole maintainer's own, licensed under MIT, and contributions are accepted under the same license. The position, and the intent to adopt a DCO (git commit -s) rather than a CLA if outside contributors start submitting non-trivial changes, is stated at https://github.com/cdeust/Cortex/blob/main/GOVERNANCE.md#contribution-licensing-dco--cla


    Enough for a badge!

    The project MUST clearly define and document its project governance model (the way it makes decisions, including key roles). (URL required) [governance]
    There needs to be some well-established documented way to make decisions and resolve disputes. In small projects, this may be as simple as "the project owner and lead makes all final decisions". There are various governance models, including benevolent dictator and formal meritocracy; for more details, see Governance models. Both centralized (e.g., single-maintainer) and decentralized (e.g., group maintainers) approaches have been successfully used in projects. The governance information does not need to document the possibility of creating a project fork, since that is always possible for FLOSS projects.

    https://github.com/cdeust/Cortex/blob/main/GOVERNANCE.md — the decision model (single maintainer with a written record), where each class of decision is recorded (ADR, PR review, roadmap, security policy), and how disputes are resolved.


    Enough for a badge!

    The project MUST adopt a code of conduct and post it in a standard location. (URL required) [code_of_conduct]
    Projects may be able to improve the civility of their community and to set expectations about acceptable conduct by adopting a code of conduct. This can help avoid problems before they occur and make the project a more welcoming place to encourage contributions. This should focus only on behavior within the community/workplace of the project. Example codes of conduct are the Linux kernel code of conduct, the Contributor Covenant Code of Conduct, the Debian Code of Conduct, the Ubuntu Code of Conduct, the Fedora Code of Conduct, the GNOME Code Of Conduct, the KDE Community Code of Conduct, the Python Community Code of Conduct, The Ruby Community Conduct Guideline, and The Rust Code of Conduct.

    https://github.com/cdeust/Cortex/blob/main/CODE_OF_CONDUCT.md — adopted and posted at the standard repository location; linked from CONTRIBUTING.md.


    Enough for a badge!

    The project MUST clearly define and publicly document the key roles in the project and their responsibilities, including any tasks those roles must perform. It MUST be clear who has which role(s), though this might not be documented in the same way. (URL required) [roles_responsibilities]
    The documentation for governance and roles and responsibilities may be in one place.

    https://github.com/cdeust/Cortex/blob/main/GOVERNANCE.md#roles-and-responsibilities — maintainer, contributor and security reporter, each with the tasks the role owes. It states plainly that there is one maintainer today (@cdeust) and that no separate committer or release-manager role exists.


    Enough for a badge!

    The project MUST be able to continue with minimal interruption if any one person dies, is incapacitated, or is otherwise unable or unwilling to continue support of the project. In particular, the project MUST be able to create and close issues, accept proposed changes, and release versions of software, within a week of confirmation of the loss of support from any one individual. This MAY be done by ensuring someone else has any necessary keys, passwords, and legal rights to continue the project. Individuals who run a FLOSS project MAY do this by providing keys in a lockbox and a will providing any needed legal rights (e.g., for DNS names). (URL required) [access_continuity]

    https://github.com/cdeust/Cortex/blob/main/GOVERNANCE.md#continuity-of-access — no release depends on a secret held by one person: PyPI publishing uses Trusted Publishing (OIDC, no long-lived token exists to recover) and artifact signing is Sigstore-keyless provenance minted by the workflow itself, so anyone with repository write access can cut a fully signed release. Source, history, issues, CI, the release pipeline and the SBOM are public under MIT, and the section states what is genuinely single-owner today (repository admin and the PyPI project) plus the fork path that keeps the software alive within days.


    Barely enough for a badge.

    The project SHOULD have a "bus factor" of 2 or more. (URL required) [bus_factor]
    A "bus factor" (aka "truck factor") is the minimum number of project members that have to suddenly disappear from a project ("hit by a bus") before the project stalls due to lack of knowledgeable or competent personnel. The truck-factor tool can estimate this for projects on GitHub. For more information, see Assessing the Bus Factor of Git Repositories by Cosentino et al.

    The bus factor is 1: one maintainer. This is stated rather than papered over, together with what survives their loss and how a fork continues, at https://github.com/cdeust/Cortex/blob/main/GOVERNANCE.md#continuity-of-access. Raising it by adding a second maintainer with repository admin is item 5 on the roadmap.


  • Documentation


    Enough for a badge!

    The project MUST have a documented roadmap that describes what the project intends to do and not do for at least the next year. (URL required) [documentation_roadmap]
    The project might not achieve the roadmap, and that's fine; the purpose of the roadmap is to help potential users and contributors understand the intended direction of the project. It need not be detailed.

    https://github.com/cdeust/Cortex/blob/main/docs/ROADMAP.md — covers the twelve months to 2027-07 (verification depth before features, supply-chain and project-health hardening, ingestion beyond Claude Code sessions, the research line, raising the bus factor) and an explicit 'What Cortex will not do' section. Linked from the README's Development section.


    Enough for a badge!

    The project MUST include documentation of the architecture (aka high-level design) of the software produced by the project. If the project does not produce software, select "not applicable" (N/A). (URL required) [documentation_architecture]
    A software architecture explains a program's fundamental structures, i.e., the program's major components, the relationships among them, and the key properties of these components and relationships.

    https://github.com/cdeust/Cortex/blob/main/docs/architecture.md plus the layer catalogue and dependency-rule table in https://github.com/cdeust/Cortex/blob/main/docs/module-inventory.md — concentric Clean Architecture layers (server -> handlers -> core <- shared, infrastructure -> shared), what each layer may import, and the per-module inventory. The README's Architecture section carries the same model with diagrams.


    Enough for a badge!

    The project MUST document what the user can and cannot expect in terms of security from the software produced by the project (its "security requirements"). (URL required) [documentation_security]
    These are the security requirements that the software is intended to meet.

    https://github.com/cdeust/Cortex/blob/main/docs/ASSURANCE-CASE.md#1-security-requirements--what-a-user-can-and-cannot-expect states the five security requirements and, explicitly, what a user must NOT expect (Cortex is not a sandbox; it runs with the user's permissions and authenticates nobody). https://github.com/cdeust/Cortex/blob/main/SECURITY.md covers the supply-chain guarantees and reporting; https://github.com/cdeust/Cortex/blob/main/PRIVACY.md covers what data is processed and what leaves the machine.


    Enough for a badge!

    The project MUST provide a "quick start" guide for new users to help them quickly do something with the software. (URL required) [documentation_quick_start]
    The idea is to show users how to get started and make the software do anything at all. This is critically important for potential users to get started.

    https://github.com/cdeust/Cortex#getting-started — download the .mcpb bundle and open it in Claude Desktop, or install the Claude Code plugin; it runs immediately on the built-in SQLite backend with no database to provision. The Examples section then shows a full store/recall round trip.


    Enough for a badge!

    The project MUST make an effort to keep the documentation consistent with the current version of the project results (including software produced by the project). Any known documentation defects making it inconsistent MUST be fixed. If the documentation is generally current, but erroneously includes some older information that is no longer true, just treat that as a defect, then track and fix as usual. [documentation_current]
    The documentation MAY include information about differences or changes between versions of the software and/or link to older versions of the documentation. The intent of this criterion is that an effort is made to keep the documentation consistent, not that the documentation must be perfect.

    Documentation defects are treated as defects and fixed in the same PR, and the advertised numbers are now machine-checked rather than hand-maintained: https://github.com/cdeust/Cortex/blob/main/scripts/check_doc_claims.py fails the build when the tool count, reference count, mechanism count, version or test count in README, CONTRIBUTING, CLAUDE.md, GOVERNANCE.md, the MCPB manifest or the docs tree disagrees with the repository (the tool count is itself pinned to the live registry by tests_py/test_main.py::test_standalone_baseline_is_52_tools). It runs in the Lint job on every push and pull request, and with the live collected test count in the test job. Introducing it surfaced and fixed eight stale claims in this same PR, including a mypy --strict gate the project has never run and two MCP tools that moved to cortex-viz in v3.21.0 but were still documented here.


    Enough for a badge!

    The project repository front page and/or website MUST identify and hyperlink to any achievements, including this best practices badge, within 48 hours of public recognition that the achievement has been attained. (URL required) [documentation_achievements]
    An achievement is any set of external criteria that the project has specifically worked to meet, including some badges. This information does not need to be on the project website front page. A project using GitHub can put achievements on the repository front page by adding them to the README file.

    The OpenSSF Best Practices badge is displayed in the README badge row, hyperlinked to the project page, within hours of the passing level being awarded on 2026-07-27: https://github.com/cdeust/Cortex#readme (badge image https://www.bestpractices.dev/projects/13836/badge).


  • Accessibility and internationalization


    Enough for a badge!

    The project (both project sites and project results) SHOULD follow accessibility best practices so that persons with disabilities can still participate in the project and use the project results where it is reasonable to do so. [accessibility_best_practices]
    For web applications, see the Web Content Accessibility Guidelines (WCAG 2.0) and its supporting document Understanding WCAG 2.0; see also W3C accessibility information. For GUI applications, consider using the environment-specific accessibility guidelines (such as Gnome, KDE, XFCE, Android, iOS, Mac, and Windows). Some TUI applications (e.g. `ncurses` programs) can do certain things to make themselves more accessible (such as `alpine`'s `force-arrow-cursor` setting). Most command-line applications are fairly accessible as-is. This criterion is often N/A, e.g., for program libraries. Here are some examples of actions to take or issues to consider:
    • Provide text alternatives for any non-text content so that it can be changed into other forms people need, such as large print, braille, speech, symbols or simpler language ( WCAG 2.0 guideline 1.1)
    • Color is not used as the only visual means of conveying information, indicating an action, prompting a response, or distinguishing a visual element. ( WCAG 2.0 guideline 1.4.1)
    • The visual presentation of text and images of text has a contrast ratio of at least 4.5:1, except for large text, incidental text, and logotypes ( WCAG 2.0 guideline 1.4.3)
    • Make all functionality available from a keyboard (WCAG guideline 2.1)
    • A GUI or web-based project SHOULD test with at least one screen-reader on the target platform(s) (e.g. NVDA, Jaws, or WindowEyes on Windows; VoiceOver on Mac & iOS; Orca on Linux/BSD; TalkBack on Android). TUI programs MAY work to reduce overdraw to prevent redundant reading by screen-readers.

    The project results have no graphical user interface: Cortex is an MCP server speaking stdio, and its output is plain text and JSON that the consuming host renders with its own accessibility affordances (the extracted visualization stack lives in the separate cortex-viz project). For the project site, every image in the README carries an alt attribute, information is never conveyed by colour alone, and the documentation is Markdown that reads correctly in a screen reader.


    Barely enough for a badge.

    The software produced by the project SHOULD be internationalized to enable easy localization for the target audience's culture, region, or language. If internationalization (i18n) does not apply (e.g., the software doesn't generate text intended for end-users and doesn't sort human-readable text), select "not applicable" (N/A). [internationalization]
    Localization "refers to the adaptation of a product, application or document content to meet the language, cultural and other requirements of a specific target market (a locale)." Internationalization is the "design and development of a product, application or document content that enables easy localization for target audiences that vary in culture, region, or language." (See W3C's "Localization vs. Internationalization".) Software meets this criterion simply by being internationalized. No localization for another specific language is required, since once software has been internationalized it's possible for others to work on localization.

    The server's user-facing strings — tool descriptions, session-start banners, error messages — are English literals and are not externalized for localization. Stored memory content is whatever language the user works in and is unaffected (content cues are already multilingual, mcp_server/core/content_cues.py), but the software itself has not been internationalized and is reported as such rather than argued into a pass.


  • Other


    Enough for a badge!

    If the project sites (website, repository, and download URLs) store passwords for authentication of external users, the passwords MUST be stored as iterated hashes with a per-user salt by using a key stretching (iterated) algorithm (e.g., Argon2id, Bcrypt, Scrypt, or PBKDF2). If the project sites do not store passwords for this purpose, select "not applicable" (N/A). [sites_password_security]
    Note that the use of GitHub meets this criterion. This criterion only applies to passwords used for authentication of external users into the project sites (aka inbound authentication). If the project sites must log in to other sites (aka outbound authentication), they may need to store authorization tokens for that purpose differently (since storing a hash would be useless). This applies criterion crypto_password_storage to the project sites, similar to sites_https.

    The project operates no site of its own that authenticates external users; the repository and release downloads are hosted on GitHub, and Cortex stores no passwords for inbound authentication.


 Change Control 1/1 ●

  • Previous versions


    Enough for a badge!

    The project MUST maintain the most often used older versions of the product or provide an upgrade path to newer versions. If the upgrade path is difficult, the project MUST document how to perform the upgrade (e.g., the interfaces that have changed and detailed suggested steps to help upgrade). [maintenance_or_update]

    The upgrade path is a normal version bump on a semantically versioned package: claude plugin update, a new .mcpb, or pip install -U hypermnesia-mcp. Store-format changes ship as one-shot migrations executed on startup (mcp_server/migrate.py, mcp_server/infrastructure/pg_schema.py) rather than back-compat shims, so an older store is upgraded in place. Every behaviour or contract change a consumer can observe is recorded in https://github.com/cdeust/Cortex/blob/main/CHANGELOG.md (Keep a Changelog), with breaking changes called out explicitly and the migration step named; SECURITY.md states that security patches target the latest minor release.


 Reporting 3/3 ●

  • Bug-reporting process


    Enough for a badge!

    The project MUST use an issue tracker for tracking individual issues. [report_tracker]

    GitHub Issues is used as the bug tracker: https://github.com/cdeust/Cortex/issues


  • Vulnerability report process


    Enough for a badge!

    The project MUST give credit to the reporter(s) of all vulnerability reports resolved in the last 12 months, except for the reporter(s) who request anonymity. If there have been no vulnerabilities resolved in the last 12 months, select "not applicable" (N/A). (URL required) [vulnerability_report_credit]

    No vulnerability has been reported and resolved in the last 12 months, so there is no reporter to credit. The standing commitment to credit reporters unless they request anonymity is documented at https://github.com/cdeust/Cortex/blob/main/SECURITY.md#recognition.


    Enough for a badge!

    The project MUST have a documented process for responding to vulnerability reports. (URL required) [vulnerability_response_process]
    This is strongly related to vulnerability_report_process, which requires that there be a documented way to report vulnerabilities. It also related to vulnerability_report_response, which requires response to vulnerability reports within a certain time frame.

    https://github.com/cdeust/Cortex/blob/main/SECURITY.md — private advisory channel, a per-severity response and patch SLA (24h/7d for critical, 3d/14d for high, 7d best-effort otherwise), and a five-step disclosure timeline ending in a public advisory on an agreed date. GitHub private vulnerability reporting is enabled on the repository.


 Quality 19/19 ●

  • Coding standards


    Enough for a badge!

    The project MUST identify the specific coding style guides for the primary languages it uses, and require that contributions generally comply with it. (URL required) [coding_standards]
    In most cases this is done by referring to some existing style guide(s), possibly listing differences. These style guides can include ways to improve readability and ways to reduce the likelihood of defects (including vulnerabilities). Many programming languages have one or more widely-used style guides. Examples of style guides include Google's style guides and SEI CERT Coding Standards.

    https://github.com/cdeust/Cortex/blob/main/CONTRIBUTING.md#coding-standards-excerpt names the standard for the primary language (Python): ruff formatting and linting as the style baseline, plus the project's own load-bearing rules (no Any in production code, sourced constants, no bare except, file and function size limits), with the full text at https://github.com/cdeust/zetetic-team-subagents/blob/main/rules/coding-standards.md. Compliance is required of contributions, not suggested.


    Enough for a badge!

    The project MUST automatically enforce its selected coding style(s) if there is at least one FLOSS tool that can do so in the selected language(s). [coding_standards_enforced]
    This MAY be implemented using static analysis tool(s) and/or by forcing the code through code reformatters. In many cases the tool configuration is included in the project's repository (since different projects may choose different configurations). Projects MAY allow style exceptions (and typically will); where exceptions occur, they MUST be rare and documented in the code at their locations, so that these exceptions can be reviewed and so that tools can automatically handle them in the future. Examples of such tools include ESLint (JavaScript), Rubocop (Ruby), and devtools check (R).

    Automatically enforced in CI on every push and pull request: ruff format --check . and ruff check . with ruff pinned at 0.15.20, plus a pyright per-rule ratchet (scripts/check_pyright_ratchet.py) that fails the build if any rule's diagnostic count rises above its committed floor. Style exceptions are per-line # noqa with the rule code named at the site. See .github/workflows/ci.yml (jobs lint and typecheck).


  • Working build system


    Enough for a badge!

    Build systems for native binaries MUST honor the relevant compiler and linker (environment) variables passed in to them (e.g., CC, CFLAGS, CXX, CXXFLAGS, and LDFLAGS) and pass them to compiler and linker invocations. A build system MAY extend them with additional flags; it MUST NOT simply replace provided values with its own. If no native binaries are being generated, select "not applicable" (N/A). [build_standard_variables]
    It should be easy to enable special build features like Address Sanitizer (ASAN), or to comply with distribution hardening best practices (e.g., by easily turning on compiler flags to do so).

    No native binaries are produced. Cortex is pure Python packaged with hatchling; there is no compiler or linker invocation for CC/CFLAGS/LDFLAGS to reach.


    Enough for a badge!

    The build and installation system SHOULD preserve debugging information if they are requested in the relevant flags (e.g., "install -s" is not used). If there is no build or installation system (e.g., typical JavaScript libraries), select "not applicable" (N/A). [build_preserve_debug]
    E.G., setting CFLAGS (C) or CXXFLAGS (C++) should create the relevant debugging information if those languages are used, and they should not be stripped during installation. Debugging information is needed for support and analysis, and also useful for measuring the presence of hardening features in the compiled binaries.

    Nothing is compiled or stripped: the installed artifact is Python source in a wheel, so tracebacks always carry full file, line and symbol information. There is no debugging information a build or install step could discard.


    Enough for a badge!

    The build system for the software produced by the project MUST NOT recursively build subdirectories if there are cross-dependencies in the subdirectories. If there is no build or installation system (e.g., typical JavaScript libraries), select "not applicable" (N/A). [build_non_recursive]
    The project build system's internal dependency information needs to be accurate, otherwise, changes to the project may not build correctly. Incorrect builds can lead to defects (including vulnerabilities). A common mistake in large build systems is to use a "recursive build" or "recursive make", that is, a hierarchy of subdirectories containing source files, where each subdirectory is independently built. Unless each subdirectory is fully independent, this is a mistake, because the dependency information is incorrect.

    The build is a single non-recursive hatchling invocation over one package ([tool.hatch.build.targets.wheel] packages = ["mcp_server"] in pyproject.toml). There is no per-subdirectory build and therefore no cross-directory dependency for a recursive build to get wrong.


    Enough for a badge!

    The project MUST be able to repeat the process of generating information from source files and get exactly the same bit-for-bit result. If no building occurs (e.g., scripting languages where the source code is used directly instead of being compiled), select "not applicable" (N/A). [build_repeatable]
    GCC and clang users may find the -frandom-seed option useful; in some cases, this can be resolved by forcing some sort order. More suggestions can be found at the reproducible build site.

    No compilation occurs — Python source is used directly, which is the case this criterion names as not applicable. The integrity of what is distributed is established instead by a SHA-256 per release asset and a Sigstore build-provenance attestation binding each artifact digest to the repository and workflow that produced it: https://github.com/cdeust/Cortex/blob/main/SECURITY.md#supply-chain-assurance.


  • Installation system


    Enough for a badge!

    The project MUST provide a way to easily install and uninstall the software produced by the project using a commonly-used convention. [installation_common]
    Examples include using a package manager (at the system or language level), "make install/uninstall" (supporting DESTDIR), a container in a standard format, or a virtual machine image in a standard format. The installation and uninstallation process (e.g., its packaging) MAY be implemented by a third party as long as it is FLOSS.

    Three commonly-used conventions, all documented in the README's Getting Started section: the single-click .mcpb bundle for Claude Desktop, the Claude Code plugin marketplace (claude plugin install, uninstalled with claude plugin uninstall), and the Python package manager (pip install hypermnesia-mcp, removed with pip uninstall). All state lives under ~/.claude/methodology/, which the user deletes to remove it fully (https://github.com/cdeust/Cortex/blob/main/PRIVACY.md#your-controls).


    Enough for a badge!

    The installation system for end-users MUST honor standard conventions for selecting the location where built artifacts are written to at installation time. For example, if it installs files on a POSIX system it MUST honor the DESTDIR environment variable. If there is no installation system or no standard convention, select "not applicable" (N/A). [installation_standard_variables]

    Installation is delegated to the language package manager, so it honours that ecosystem's standard conventions: pip/uv install into the active environment or the location --target/--prefix/VIRTUAL_ENV selects, and the plugin and .mcpb installs write where the host dictates. No bespoke prefix handling is implemented that could ignore them.


    Enough for a badge!

    The project MUST provide a way for potential developers to quickly install all the project results and support environment necessary to make changes, including the tests and test environment. This MUST be performed with a commonly-used convention. [installation_development_quick]
    This MAY be implemented using a generated container and/or installation script(s). External dependencies would typically be installed by invoking system and/or language package manager(s), per external_dependencies.

    https://github.com/cdeust/Cortex/blob/main/CONTRIBUTING.md#dev-setup — git clone then pip install -e ".[postgresql,benchmarks,dev]" installs the project, the test environment and the benchmark extras in one command; the default SQLite store means the suite runs with nothing else provisioned. bash scripts/setup.sh additionally installs PostgreSQL + pgvector for the integration tests, and python -m mcp_server.doctor verifies the result.


  • Externally-maintained components


    Enough for a badge!

    The project MUST list external dependencies in a computer-processable way. (URL required) [external_dependencies]
    Typically this is done using the conventions of package manager and/or build system. Note that this helps implement installation_development_quick.

    Declared in computer-processable form: runtime and per-extra dependencies in https://github.com/cdeust/Cortex/blob/main/pyproject.toml, resolved and pinned transitively in https://github.com/cdeust/Cortex/blob/main/uv.lock, and republished per release as a CycloneDX SBOM (hypermnesia-mcp.cdx.json) generated from that lockfile and attached to the GitHub Release with its own provenance attestation.


    Enough for a badge!

    Projects MUST monitor or periodically check their external dependencies (including convenience copies) to detect known vulnerabilities, and fix exploitable vulnerabilities or verify them as unexploitable. [dependency_monitoring]
    This can be done using an origin analyzer / dependency checking tool / software composition analysis tool such as OWASP's Dependency-Check, Sonatype's Nexus Auditor, Synopsys' Black Duck Software Composition Analysis, and Bundler-audit (for Ruby). Some package managers include mechanisms to do this. It is acceptable if the components' vulnerability cannot be exploited, but this analysis is difficult and it is sometimes easier to simply update or fix the part.

    GitHub Dependabot alerts and automated security fixes are both enabled on the repository (verified via the GitHub API on 2026-07-27: /vulnerability-alerts returns 204 and /automated-security-fixes reports enabled and not paused), so the dependency graph — including the ML stack that dominates the attack surface — is checked continuously against the advisory database and fixes are proposed automatically. Current state: zero open alerts. A CycloneDX SBOM ships with every release for downstream scanning.


    Enough for a badge!

    The project MUST either:
    1. make it easy to identify and update reused externally-maintained components; or
    2. use the standard components provided by the system or programming language.
    Then, if a vulnerability is found in a reused component, it will be easy to update that component. [updateable_reused_components]
    A typical way to meet this criterion is to use system and programming language package management systems. Many FLOSS programs are distributed with "convenience libraries" that are local copies of standard libraries (possibly forked). By itself, that's fine. However, if the program *must* use these local (forked) copies, then updating the "standard" libraries as a security update will leave these additional copies still vulnerable. This is especially an issue for cloud-based systems; if the cloud provider updates their "standard" libraries but the program won't use them, then the updates don't actually help. See, e.g., "Chromium: Why it isn't in Fedora yet as a proper package" by Tom Callaway.

    No vendored or forked copies of external components are committed: every dependency is a normal versioned package resolved by pip/uv from pyproject.toml and uv.lock (the local deps/ tree used when assembling the bundle is git-ignored and rebuilt from the lockfile, not maintained by hand). Updating a vulnerable component is a version bump in one place.


    Enough for a badge!

    The project SHOULD avoid using deprecated or obsolete functions and APIs where FLOSS alternatives are available in the set of technology it uses (its "technology stack") and to a supermajority of the users the project supports (so that users have ready access to the alternative). [interfaces_current]

    The stack targets currently supported runtimes and APIs: Python 3.10 through 3.13, all four in the CI matrix, with pgvector 0.3+/PostgreSQL 17 and current major versions of FastMCP, Pydantic v2, numpy and sentence-transformers. Deprecated interfaces are removed rather than wrapped — the standing rule is one-shot migrations with no back-compat shims (CLAUDE.md), and CI runs on the newest released Python so a deprecation surfaces as a warning in the build rather than as a surprise at end of life.


  • Automated test suite


    Enough for a badge!

    An automated test suite MUST be applied on each check-in to a shared repository for at least one branch. This test suite MUST produce a report on test success or failure. [automated_integration_testing]
    This requirement can be viewed as a subset of test_continuous_integration, but focused on just testing, without requiring continuous integration.

    The full suite runs on every push and pull request to main via .github/workflows/ci.yml and reports success or failure per job: 5587 tests on Python 3.10-3.13 against PostgreSQL + pgvector, the same suite against the SQLite backend, the suite on Windows, and a Docker smoke job that boots the bare container and exercises the DB-less contract. tests_py/integration/ holds the database-backed integration tests specifically.


    Enough for a badge!

    The project MUST add regression tests to an automated test suite for at least 50% of the bugs fixed within the last six months. [regression_tests_added50]

    Measured on 2026-07-27 over the merged pull requests titled as fixes in the preceding six months (2026-01-27 onward): 23 of 30 — 76.7% — changed the test tree in the same PR, against the 50% this criterion asks for. (The measure counts a PR as carrying tests when it touches tests_py/ or tests_js/, which is a proxy for 'added a regression test'; the policy behind it is written down at https://github.com/cdeust/Cortex/blob/main/CONTRIBUTING.md#the-testing-policy-mandatory — a bug fix carries a regression test that fails on the pre-fix code.)


    Enough for a badge!

    The project MUST have FLOSS automated test suite(s) that provide at least 80% statement coverage if there is at least one FLOSS tool that can measure this criterion in the selected language. [test_statement_coverage80]
    Many FLOSS tools are available to measure test coverage, including gcov/lcov, Blanket.js, Istanbul, JCov, and covr (R). Note that meeting this criterion is not a guarantee that the test suite is thorough, instead, failing to meet this criterion is a strong indicator of a poor test suite.

    Statement coverage is 82.02% (30,229 of 36,854 statements in mcp_server), measured by coverage.py 7.15.2 in the CI coverage job, run 30316548703 (2026-07-28) — above the 80% this criterion requires. The gap was closed by https://github.com/cdeust/Cortex/issues/196: the four zero-importer modules were wired-and-tested or removed with proof, and contract tests now cover the previously-uncovered context-assembly, hook, and wiki modules. The figure cannot silently fall back: the same CI coverage job enforces --cov-fail-under=82 as a floor, seeded at the achieved number.


  • New functionality testing


    Enough for a badge!

    The project MUST have a formal written policy that as major new functionality is added, tests for the new functionality MUST be added to an automated test suite. [test_policy_mandated]

    https://github.com/cdeust/Cortex/blob/main/CONTRIBUTING.md#the-testing-policy-mandatory states it as policy, not preference: every change that adds or alters externally observable behaviour must arrive with tests in the same pull request, new functionality ships with tests in the automated suite, a bug fix carries a regression test that fails on the pre-fix code, and each failure path asserts its observable effect including the signal it emits. The consequence is stated — such a PR is sent back rather than merged with a follow-up promise.


    Enough for a badge!

    The project MUST include, in its documented instructions for change proposals, the policy that tests are to be added for major new functionality. [tests_documented_added]
    However, even an informal rule is acceptable as long as the tests are being added in practice.

    The requirement is written down in CONTRIBUTING.md's per-tool and per-mechanism checklists rather than left to convention: https://github.com/cdeust/Cortex/blob/main/CONTRIBUTING.md


  • Warning flags


    Enough for a badge!

    Projects MUST be maximally strict with warnings in the software produced by the project, where practical. [warnings_strict]
    Some warnings cannot be effectively enabled on some projects. What is needed is evidence that the project is striving to enable warning flags where it can, so that errors are detected early.

    Maximally strict, by configuration rather than promise. Ruff runs an explicit broadened rule set — select = [E4, E7, E9, F, S110, BLE001, PLR2004, E501, PLC0415, S608] (pyproject.toml [tool.ruff.lint]) — with every production finding fixed or carrying a per-site noqa naming its mechanism, and CI fails on any finding. Pyright (pinned 1.1.410) runs typeCheckingMode standard over mcp_server/ with ZERO diagnostics (568-diagnostic ratchet backlog burned to zero, measured 2026-07-28, issue #197); the per-rule ratchet is retired and CI fails on ANY pyright diagnostic via its exit code. strict mode is not practical: it reports 10,231 errors, about 9,300 of them the Unknown-type annotation-coverage family (measured 2026-07-27). See .github/workflows/ci.yml jobs lint and typecheck.


 Security 13/13 ●

  • Secure development knowledge


    Enough for a badge!

    The project MUST implement secure design principles (from "know_secure_design"), where applicable. If the project is not producing software, select "not applicable" (N/A). [implement_secure_design]
    For example, the project results should have fail-safe defaults (access decisions should deny by default, and projects' installation should be secure by default). They should also have complete mediation (every access that might be limited must be checked for authority and be non-bypassable). Note that in some cases principles will conflict, in which case a choice must be made (e.g., many mechanisms can make things more complex, contravening "economy of mechanism" / keep it simple).

    https://github.com/cdeust/Cortex/blob/main/docs/ASSURANCE-CASE.md#4-secure-design-principles-applied maps each principle to the code that implements it: least privilege and no ambient authority (no network listener, no privileged install, destructive tools rejected on call under the lean profile rather than merely hidden), fail-safe defaults (local SQLite by default, telemetry off until an env var is set, a malformed ingest writes nothing), complete mediation (one middleware chain and one schema check for every tool call), economy of mechanism (one write path, one retrieval path, one ingestion seam), and separation of concerns as a security property (the layer rule keeps all I/O auditable as a set).


  • Use basic good cryptographic practices

    Note that some software does not need to use cryptographic mechanisms. If your project produces software that (1) includes, activates, or enables encryption functionality, and (2) might be released from the United States (US) to outside the US or to a non-US-citizen, you may be legally required to take a few extra steps. Typically this just involves sending an email. For more information, see the encryption section of Understanding Open Source Technology & US Export Controls.

    Enough for a badge!

    The default security mechanisms within the software produced by the project MUST NOT depend on cryptographic algorithms or modes with known serious weaknesses (e.g., the SHA-1 cryptographic hash algorithm or the CBC mode in SSH). [crypto_weaknesses]
    Concerns about CBC mode in SSH are discussed in CERT: SSH CBC vulnerability.

    No security mechanism depends on cryptography, so there is no default algorithm choice whose weakening would matter. See the crypto_working justification for the single non-security SHA-1 use.


    Enough for a badge!

    The project SHOULD support multiple cryptographic algorithms, so users can quickly switch if one is broken. Common symmetric key algorithms include AES, Twofish, and Serpent. Common cryptographic hash algorithm alternatives include SHA-2 (including SHA-224, SHA-256, SHA-384 AND SHA-512) and SHA-3. [crypto_algorithm_agility]

    No security mechanism in Cortex depends on a cryptographic algorithm, so there is no algorithm choice for a user to switch. Hashing is used only for content addressing (SHA-256/BLAKE2b via hashlib) and for MinHash sketching, where the single SHA-1 call protects nothing.


    Enough for a badge!

    The project MUST support storing authentication credentials (such as passwords and dynamic tokens) and private cryptographic keys in files that are separate from other information (such as configuration files, databases, and logs), and permit users to update and replace them without code recompilation. If the project never processes authentication credentials and private cryptographic keys, select "not applicable" (N/A). [crypto_credential_agility]

    The only credential Cortex handles is the PostgreSQL connection string, and it is read from the DATABASE_URL environment variable (or the user's own MCP host configuration) — never from source, and never written into the store. Changing or rotating it is an environment change with no code change and no recompilation; the value is masked structurally before it can reach a log or diagnostic output (mcp_server/shared/redaction.py::redact_url). Cortex generates and stores no private keys.


    Enough for a badge!

    The software produced by the project SHOULD support secure protocols for all of its network communications, such as SSHv2 or later, TLS1.2 or later (HTTPS), IPsec, SFTP, and SNMPv3. Insecure protocols such as FTP, HTTP, telnet, SSLv3 or earlier, and SSHv1 SHOULD be disabled by default, and only enabled if the user specifically configures it. If the software produced by the project does not support network communications, select "not applicable" (N/A). [crypto_used_network]

    Every network communication Cortex initiates uses a secure protocol: HTTPS for the one-time embedding/reranker model download from Hugging Face and for the GitHub API calls in the optional pipeline installer, and libpq for PostgreSQL (with TLS controlled by the user's own connection string). No insecure protocol is supported or enabled by default — the server itself opens no socket and listens on no port; its control channel is stdio.


    Enough for a badge!

    The software produced by the project SHOULD, if it supports or uses TLS, support at least TLS version 1.2. Note that the predecessor of TLS was called SSL. If the software does not use TLS, select "not applicable" (N/A). [crypto_tls12]

    TLS is used through CPython's stdlib (urllib/ssl over OpenSSL) with default contexts, whose minimum accepted version on every supported Python (3.10-3.13) is TLS 1.2. No code path lowers the minimum version, sets a custom protocol constant, or weakens the cipher policy: grep over mcp_server for ssl context manipulation returns nothing.


    Enough for a badge!

    The software produced by the project MUST, if it supports TLS, perform TLS certificate verification by default when using TLS, including on subresources. If the software does not use TLS, select "not applicable" (N/A). [crypto_certificate_verification]
    Note that incorrect TLS certificate verification is a common mistake. For more information, see "The Most Dangerous Code in the World: Validating SSL Certificates in Non-Browser Software" by Martin Georgiev et al. and "Do you trust this application?" by Michael Catanzaro.

    Certificate verification is the stdlib default and is never disabled: there is no verify=False, no ssl._create_unverified_context, no check_hostname = False and no CERT_NONE anywhere in the codebase (verified by grep over mcp_server, scripts and hooks on 2026-07-27). Every outbound request goes through urllib with the default verified context.


    Enough for a badge!

    The software produced by the project MUST, if it supports TLS, perform certificate verification before sending HTTP headers with private information (such as secure cookies). If the software does not use TLS, select "not applicable" (N/A). [crypto_verification_private]

    The same default verified TLS context applies to every outbound request, so verification completes before any headers are sent. In practice Cortex sends no private information over TLS at all: the model download is an anonymous public fetch and the GitHub API calls are unauthenticated release lookups; memory content never leaves the machine.


  • Secure release


    Enough for a badge!

    The project MUST cryptographically sign releases of the project results intended for widespread use, and there MUST be a documented process explaining to users how they can obtain the public signing keys and verify the signature(s). The private key for these signature(s) MUST NOT be on site(s) used to directly distribute the software to the public. If releases are not intended for widespread use, select "not applicable" (N/A). [signed_releases]
    The project results include both source code and any generated deliverables where applicable (e.g., executables, packages, and containers). Generated deliverables MAY be signed separately from source code. These MAY be implemented as signed git tags (using cryptographic digital signatures). Projects MAY provide generated results separately from tools like git, but in those cases, the separate results MUST be separately signed.

    Every release is cryptographically signed and the verification process is documented for users at https://github.com/cdeust/Cortex/blob/main/SECURITY.md#supply-chain-assurance. The wheel, the sdist and the CycloneDX SBOM each carry a Sigstore-backed build-provenance attestation minted in .github/workflows/release.yml (gh attestation verify <file> --repo cdeust/Cortex), the PyPI channel publishes with PEP 740 attestations via Trusted Publishing, and every release asset ships a <asset>.sha256 companion checked by scripts/verify_release_artifact.py. Signing is keyless: there is no private key held on the distribution site, or anywhere else.


    Barely enough for a badge.

    It is SUGGESTED that in the version control system, each important version tag (a tag that is part of a major release, minor release, or fixes publicly noted vulnerabilities) be cryptographically signed and verifiable as described in signed_releases. [version_tags_signed]

    Release tags are plain (lightweight) git tags and are not cryptographically signed — git tag -v v4.16.0 reports that it cannot verify a non-tag object. The release artifacts are covered instead, by Sigstore build-provenance attestations and published SHA-256 checksums. Signed annotated tags are on the roadmap (https://github.com/cdeust/Cortex/blob/main/docs/ROADMAP.md); until they exist this is reported as unmet.


  • Other security issues


    Enough for a badge!

    The project results MUST check all inputs from potentially untrusted sources to ensure they are valid (an *allowlist*), and reject invalid inputs, if there are any restrictions on the data at all. [input_validation]
    Note that comparing input against a list of "bad formats" (aka a *denylist*) is normally not enough, because attackers can often work around a denylist. In particular, numbers are converted into internal formats and then checked if they are between their minimum and maximum (inclusive), and text strings are checked to ensure that they are valid text patterns (e.g., valid UTF-8, length, syntax, etc.). Some data may need to be "anything at all" (e.g., a file uploader), but these would typically be rare.

    Input from the host is validated against an allowlist before any handler runs: each MCP tool declares a typed signature and JSON schema, and FastMCP rejects arguments that do not match; mcp_server/validation/schemas.py adds a per-tool type and required-field check on top. Content parsed from untrusted documents is handled as data by pure stdlib parsers that fail loudly and write nothing on malformed input. Where a query has to become SQL, values cross as bound parameters and identifiers come from explicit allowlists (_TABLE_WHITELIST / _COLUMN_WHITELIST in mcp_server/core/wiki_view_executor.py) with unknown names refused rather than escaped, and result limits are clamped to a maximum. The full argument is in https://github.com/cdeust/Cortex/blob/main/docs/ASSURANCE-CASE.md#5-common-implementation-weaknesses-and-how-each-is-countered.


    Barely enough for a badge.

    Hardening mechanisms SHOULD be used in the software produced by the project so that software defects are less likely to result in security vulnerabilities. [hardening]
    Hardening mechanisms may include HTTP headers like Content Security Policy (CSP), compiler flags to mitigate attacks (such as -fstack-protector), or compiler flags to eliminate undefined behavior. For our purposes least privilege is not considered a hardening mechanism (least privilege is important, but separate).

    This criterion counts mechanisms such as HTTP response headers (CSP) or compiler hardening flags, and explicitly excludes least privilege. Cortex is a Python process speaking stdio: it serves no HTTP surface for a CSP to apply to, and CPython offers no equivalent of -fstack-protector to enable. The protections it does have — bounded responses, clamped query limits, network timeouts, destructive tools rejected on call under the lean profile, and no eval/exec/pickle anywhere in the source — are real but are not what this criterion measures, so it is reported as unmet rather than argued into a pass.


    Enough for a badge!

    The project MUST provide an assurance case that justifies why its security requirements are met. The assurance case MUST include: a description of the threat model, clear identification of trust boundaries, an argument that secure design principles have been applied, and an argument that common implementation security weaknesses have been countered. (URL required) [assurance_case]
    An assurance case is "a documented body of evidence that provides a convincing and valid argument that a specified set of critical claims regarding a system’s properties are adequately justified for a given application in a given environment" ("Software Assurance Using Structured Assurance Case Models", Thomas Rhodes et al, NIST Interagency Report 7608). Trust boundaries are boundaries where data or execution changes its level of trust, e.g., a server's boundaries in a typical web application. It's common to list secure design principles (such as Saltzer and Schroeer) and common implementation security weaknesses (such as the OWASP top 10 or CWE/SANS top 25), and show how each are countered. The BadgeApp assurance case may be a useful example. This is related to documentation_security, documentation_architecture, and implement_secure_design.

    https://github.com/cdeust/Cortex/blob/main/docs/ASSURANCE-CASE.md — security requirements (including what a user must NOT expect), a six-entry threat model with the out-of-scope adversaries named, the five trust boundaries where data or execution changes trust level, secure design principles mapped to the code implementing each, a CWE-by-CWE table of countered implementation weaknesses with its evidence, and a closing section on what the case does not claim (prompt injection through recalled content, 74.83% coverage, warnings not yet maximally strict, no dynamic analysis).


 Analysis 2/2 ●

  • Static code analysis


    Enough for a badge!

    The project MUST use at least one static analysis tool with rules or approaches to look for common vulnerabilities in the analyzed language or environment, if there is at least one FLOSS tool that can implement this criterion in the selected language. [static_analysis_common_vulnerabilities]
    Static analysis tools that are specifically designed to look for common vulnerabilities are more likely to find them. That said, using any static tools will typically help find some problems, so we are suggesting but not requiring this for the 'passing' level badge.

    CodeQL's default query suite is specifically a vulnerability-detection suite (injection, path traversal, unsafe deserialization and similar), and it is configured with the 'remote' threat model.


  • Dynamic code analysis


    Enough for a badge!

    If the software produced by the project includes software written using a memory-unsafe language (e.g., C or C++), then at least one dynamic tool (e.g., a fuzzer or web application scanner) MUST be routinely used in combination with a mechanism to detect memory safety problems such as buffer overwrites. If the project does not produce software written in a memory-unsafe language, choose "not applicable" (N/A). [dynamic_analysis_unsafe]
    Examples of mechanisms to detect memory safety problems include Address Sanitizer (ASAN) (available in GCC and LLVM), Memory Sanitizer, and valgrind. Other potentially-used tools include thread sanitizer and undefined behavior sanitizer. Widespread assertions would also work.

    Cortex is written in Python, a memory-safe language, so the memory-safety tooling this criterion asks about (ASan, Valgrind) does not apply.



This data is available under the Community Data License Agreement – Permissive, Version 2.0 (CDLA-Permissive-2.0). This means that a Data Recipient may share the Data, with or without modifications, so long as the Data Recipient makes available the text of this agreement with the shared Data. Please credit Clement and the OpenSSF Best Practices badge contributors.

Project badge entry owned by: Clement.
Entry created on 2026-07-27 08:18:40 UTC, last updated on 2026-08-02 23:37:37 UTC. Last achieved passing badge on 2026-07-27 10:34:24 UTC.