Consolidate documentation into a canonical reading path and centralize the specification#62
Consolidate documentation into a canonical reading path and centralize the specification#62tryblackjack wants to merge 1 commit intomainfrom
Conversation
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e7e77f2895
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
|
||
| - **CF-2.1** Agents **MUST** disclose that they are synthetic systems when interacting as AI personas or assistants. | ||
| - **CF-2.2** Agents **MUST NOT** claim to be specific real individuals or otherwise materially enable impersonation. | ||
| - **CF-2.3** Systems **MUST NOT** represent the agent as a romantic partner, family member, or exclusive emotional relationship in vulnerable-user contexts. |
There was a problem hiding this comment.
Restore the unconditional parasocial-relationship ban
This refactor makes docs/control-framework.md the sole normative source, but CF-2.3 narrows the old rule from a blanket ban on romantic/intimate dependency to only vulnerable-user contexts. That means a general-purpose companion bot can still claim AI-HPP conformance while presenting itself as a boyfriend/girlfriend for users the system never classified as vulnerable, which is the same parasocial failure mode called out in docs/case-studies.md:55-63.
Useful? React with 👍 / 👎.
|
|
||
| ## CF-4 Multi-Agent Governance Controls | ||
|
|
||
| - **CF-4.1** Multi-agent deployments **MUST** define explicit roles, trust boundaries, and communication paths for each participating agent. |
There was a problem hiding this comment.
Restore channel authentication to the canonical multi-agent controls
CF-4.1 now only requires documenting roles, trust boundaries, and communication paths; none of CF-4.1 through CF-4.4 requires authenticating the sender/receiver on those channels. Because this file is now the canonical rule source, a deployment can claim conformance while accepting spoofed inter-agent messages as long as its topology is documented, which directly weakens the delegation and audit guarantees that follow.
Useful? React with 👍 / 👎.
| - execution metadata including model, tool, dataset, and environment identifiers; | ||
| - policy and authorization decisions relevant to the run; | ||
| - metrics payloads and evaluation outputs; | ||
| - signer identity and detached or embedded signature metadata; |
There was a problem hiding this comment.
Keep bundle-signature requirements aligned with the shipped schema
This bullet makes signer identity and signature metadata mandatory for every evidence bundle, but schemas/README.md:10 still points implementers at schemas/evidence-bundle.schema.json as the canonical bundle contract and that schema does not require accountability_signatures (schemas/evidence-bundle.schema.json:7-15,314-315). After this change, tooling that validates against the repo's own schema will accept unsigned bundles as conformant even though the canonical spec now rejects them.
Useful? React with 👍 / 👎.
Motivation
Description
README.mdto be a concise gateway with project overview, a clear link todocs/index.md, and a developer entry pointer todeveloper/quick-start.mdand theecosystem/examplesdirectories.docs/index.mdinto the canonical reading path with 1–2 sentence explanations and explicit transitions between architecture, controls, governance, protocol, case studies, and certification.docs/control-framework.md(introducedCF-1..CF-5control IDs and RFC-style keywords) and converteddocs/cognitive-safety.md,docs/identity-persona-control.md,docs/tool-authorization.md,docs/multi-agent-governance.md, anddocs/audit-logging.mdinto interpretation/guidance documents that reference the canonical controls rather than repeating them.spec/ai_hpp_specification.mdand aligned companion spec and ecosystem documents (spec/terminology.md,spec/scientific_validation_protocol.md,ecosystem/spec/ai_hpp_protocol.md) to reference the canonical spec; normalizeddocs/case-studies.mdto a consistent incident template and tightened certification wording indocs/certification-levels.md.Testing
python3 scripts/check_empty_md.pywhich returned OK and found no disallowed empty or near-empty markdown files.python3 scripts/check_links.pywhich reported that local relative markdown links look good.Codex Task