Skip to content

feat(skills): expose bounded search and read tools#260

Open
andrei-hasna wants to merge 1 commit into
mainfrom
feat/model-facing-skill-tools-OPE2-00068
Open

feat(skills): expose bounded search and read tools#260
andrei-hasna wants to merge 1 commit into
mainfrom
feat/model-facing-skill-tools-OPE2-00068

Conversation

@andrei-hasna

Copy link
Copy Markdown
Contributor

Summary

  • expose namespaced skills.search and skills.read tools for turn-available skill packages
  • preserve exact listing-provider provenance across opaque authority, package, and resource handles
  • bound arguments, inspected results, serialized output, provider errors, lifecycle state, and telemetry previews
  • add integration coverage for same-kind routing, custom authorities, host reads, hostile bounds, resource mismatch, and stale turns

Verification

  • just test-fast -p codex-skills-extension
  • just test -p codex-skills-extension
  • just fix -p codex-skills-extension
  • just fmt
  • just bazel-lock-update
  • just bazel-lock-check
  • staged and committed-range secret and Co-Authored-By scans
  • two independent adversarial reviews: no P0 or P1 findings

Station02 lacks OpenSSL development headers, so Rust compilation used a temporary vendored OpenSSL feature that was removed from the committed manifest and lockfile. CI should exercise the normal repository dependency configuration.

Residual risk

  • the legacy public SkillProviders search seam remains kind-routed; these model tools use exact turn-scoped routes
  • providers can allocate large result vectors before the model-tool boundary applies processing and output caps

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.

1 participant