Skip to content

Harden contribution gate, align README counts, add signed trust attes… - #8

Merged
criptogus merged 1 commit into
mainfrom
claude/audit-project-improvements-wm0w8
May 16, 2026
Merged

Harden contribution gate, align README counts, add signed trust attes…#8
criptogus merged 1 commit into
mainfrom
claude/audit-project-improvements-wm0w8

Conversation

@criptogus

Copy link
Copy Markdown
Owner

…tations

  • validate:content now runs the prompt-injection guard on every package: critical signals (fake SYSTEM/control tokens, secret-exfil URLs) fail the build; high/medium are non-fatal warnings (legit skill prose trips the heuristic), with x_security_research opt-out. Closes the PR/git hole that the MCP upload path already covered.
  • README "What's inside" now distinguishes the open-seed counts in this repo (76/2/6/2) from the hosted-registry counts, instead of overstating.
  • New scripts/generate-trust-attestation.mjs + verify-trust-attestation.mjs: Ed25519-signed, dependency-free, offline-verifiable attestation binding a package's content hash + version to its adversarial result, so a security team can confirm a Trust Score without trusting the hosted endpoint.

https://claude.ai/code/session_01CV6zb1KBVoe3eBttyK9U4Z

What does this PR add?

Type

  • New package (skill / playbook / soul / guardrail)
  • Improvement to an existing package (bumped version)
  • Platform code or docs

Checklist (for content PRs)

  • Filename matches slug and the type's folder
  • At least 2 worked examples (skills) — realistic input + exact expected output
  • Original work, public-domain, or properly attributed; no secrets / PII
  • bun run validate:content passes locally
  • I've read CONTRIBUTING.md and agree to license under Apache-2.0 (code) / CC-BY-SA-4.0 (content)

Notes for reviewers

…tations

- validate:content now runs the prompt-injection guard on every package:
  critical signals (fake SYSTEM/control tokens, secret-exfil URLs) fail the
  build; high/medium are non-fatal warnings (legit skill prose trips the
  heuristic), with x_security_research opt-out. Closes the PR/git hole that
  the MCP upload path already covered.
- README "What's inside" now distinguishes the open-seed counts in this repo
  (76/2/6/2) from the hosted-registry counts, instead of overstating.
- New scripts/generate-trust-attestation.mjs + verify-trust-attestation.mjs:
  Ed25519-signed, dependency-free, offline-verifiable attestation binding a
  package's content hash + version to its adversarial result, so a security
  team can confirm a Trust Score without trusting the hosted endpoint.

https://claude.ai/code/session_01CV6zb1KBVoe3eBttyK9U4Z
@criptogus
criptogus merged commit 7c0b3de into main May 16, 2026
1 of 2 checks passed
criptogus pushed a commit that referenced this pull request May 28, 2026
…equest

Infra/API follow-ups from operator feedback:

- review_skills_batch: score up to 10 files in one parallel call, aligned to
  input order with a single shared feedback_request — far lower round-trip
  latency than N review_skill calls (feedback #8).
- In-process result cache keyed by content_hash + options. Identical
  re-submissions skip both the deterministic scoring and the LLM semantic pass,
  cutting ~15-30s to ~0 on iterative loops; responses carry `cached` (#9).
  Core review logic extracted into computeReview(), shared by single + batch.
- dry_run preview for upload_packages and request_primitive: validates the file
  (type inference + prompt-injection guard) or echoes the would-be request
  WITHOUT auth, persistence, or model budget, so agents can test the publish
  flow before connecting OAuth. The MCP write-tool auth gate now lets a
  dry_run call through anonymously (#5/#6).

https://claude.ai/code/session_01EntkmBiYh381pKqvvSFBkg
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants