Repository navigation
Conversation
|
نسخة مكررة من |
The stable Skills page missed the final SEP's Agent Skills compatibility clause, and its examples used abbreviated hashes and inconsistent byte sizes. Archived drafts also retained current-sounding source-of-truth guidance. This corrects the stable page directly to match already-established requirements: restores the compatibility clause, shows the resources capability and server/discover, explains omitted request metadata, and makes the example manifests and content consistent. It also labels archived documents as historical and fixes the SEP domain in AGENTS.md. This is an editorial alignment, not a new extension proposal; #142 remains separate. Validation: all JSON examples parsed; all 13 manifest entries checked against sample bytes for SHA-256 and size; the corrected specification builds with the VitePress publishing site; git diff --check passed. AI assistance: prepared with Codex.
|
Field support for this, plus one gap it leaves open. The executable-content line (fetched before approval, hashed by the host, retained immutably, executed only if the retained bytes match) is exactly the loop we run in production for skill files (rugsnare, open source): raw-octet sha-256 per file, recorded at approval, compared on every subsequent load - the digest is deliberately sha256sum-compatible so a host can verify outside our tool. From operating it, two hash regimes matter, and the spec's "hashed" deserves both:
A companion line worth considering: "Hosts SHOULD detect changes to previously approved instructional content and surface them to the user before the changed content takes effect." That is the missing half of approval-to-load: an approval that is silent about later mutation only covers the first load. Implementation reference (Apache-2.0): https://github.com/Paraphern/rugsnare - per-file raw-octet digests for skills, canonicalized schema/prose split for MCP tool contracts, was/became diff reporting. |
|
gentle reminder. |
Motivation and Context
Improves security
Breaking Changes
Types of changes
Checklist
Additional context