Skip to content

session-peer skill v0.5.0

Choose a tag to compare

@abruption abruption released this 02 Nov 04:57
· 2 commits to main since this release
ee2d16e

Python skill 0.5.0 documents the optional MCP workflows and structured CLI preview in published Python runtime v1.4.0. Runtime and skill versions are independently managed.

Changes

  • Add policy metadata discovery, approved preset selection, explicit preflight and bounded diagnostics alongside existing authorized sends.
  • Separate policy schema 2 permissions, host-client approvals and unavailable caller identity. Metadata lookup does not grant send authority, readiness or ACK.
  • Scope ordinary CLI JSON separately from host-free structured/MCP projections, preserving text-only --explain, successful legacy list/send payloads and their limits.
  • Preserve unknown/refusal handling, known submission evidence, no automatic resend and the distinction between queueing/submission and consumption/ACK.
  • Align English, Korean, Japanese and Simplified Chinese usage/README guidance and keep historical feature floors/reference links.

Compatibility

  • Skill version: 0.5.0
  • Basic CLI minimum: 0.9.1
  • Published CLI coverage reference: 1.4.0
  • Receiver codexBin feature floor: 1.0.1

Optional features still require their runtime/help, Python extras, platform, policy, permission and native-writer guards. Experimental handoff remains outside baseline/full coverage. The independent TypeScript skill profile is unchanged.

Verification and limits

Runtime source was checked against 476c67e636ef02cb93797dd2f42252dfe577625f, including the final timeout/preflight defenses. Local validation passed 41 coverage regressions, metadata/tag consistency, frontmatter validation, published-source anchors, localized identifiers and isolated Skills CLI discovery/installation. These are document/package checks, not new live native/SSH/ACK, model or operations qualification.

Read-only translation assistance used Antigravity CLI 1.4.12 / Gemini 3.8 Flash Medium, medium effort, on candidate MCP sections/tables and README notices across eight files (194.849 seconds). Codex validated findings and language fixes. Later published-source/status/version finalization and the MCP-only scope of server/policy prerequisites were checked by Codex/static source review, not another external-model run.

Runtime publication does not install this independently managed skill. Use the existing skill manager and an explicitly chosen published release. A Python compatibility snapshot follows separately from the exact published skill tag/source/hash; existing runtime/skill assets, user installations and operating services are preserved.

Exact source and CI

  • PR: #43, reviewed head be3a7d24d9115118183e3f5d506901e7a78de335
  • Annotated tag: v0.5.0, object fda9c6421c24eb8a4f58f6125e5f703c92252e1e
  • Peeled source: ee2d16e954b6a0204f9a085ad3c308f4fba83ccd (tree identical to the reviewed PR)
  • Canonical SKILL: 29,326 UTF-8 bytes, SHA-256 532d370c3acf18f59d2b53213bfa9b08a877fcb3415de2f3fb607219c665ca0d
  • PR CI and merge-main CI: success

Automatic code and security reviews both completed on the PR head, with no posted findings and a review-bot approval reaction. Code-review finding flags are false; the security summary's finding flags are null, so no stronger machine-readable security finding assertion is made.

Exact-tag CI passed on the peeled release source before publication.