openspec-update-change
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as user-provided notes, project-level glossary files, and existing specifications to generate documentation revisions.
- Ingestion points: Reads project-level and change-level
GLOSSARY.md,VERIFY.md,SPEC.md, and user-referenced notes/conversation context in Steps 2, 3, and 4. - Boundary markers: No explicit delimiter markers or "ignore embedded instructions" directives are specified for the ingested external content.
- Capability inventory: The skill has the capability to write revisions to planning artifacts, glossaries, and verification plans as defined in
artifactPaths(Step 5). - Sanitization: Risk is mitigated by a mandatory human-in-the-loop confirmation step before any file is written and the strict enforcement of Simplified Technical English (STE) for all generated content.
- [COMMAND_EXECUTION]: The skill invokes the
openspecCLI to query change status and retrieval instructions. These operations are limited to metadata and state retrieval (e.g.,list,status,instructions) and do not perform system modifications. - [SAFE]: The instructions include explicit prohibitions against editing implementation code, creating extension files, or running verification commands, ensuring the skill stays within its documentation-focused scope.
Audit Metadata