sync-skills-shared-protocols
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill workflow involves executing several local Python and Node.js scripts to manage protocol synchronization.
- Evidence: Instructions call for running ".claude/scripts/sync-update-blocks.py", ".claude/scripts/build-protocol-projection.cjs", and ".claude/scripts/codex/verify-sync-adoption-parity.mjs".
- [INDIRECT_PROMPT_INJECTION]: The skill acts as a content delivery mechanism, copying instructions from a central source file into multiple other files. This creates a surface where a single change to the source file is automatically replicated across all associated agents and skills.
- Ingestion points: The skill reads content from the canonical source file ".claude/skills/shared/sync-inline-versions.md".
- Boundary markers: The skill uses explicit "SYNC" and "/SYNC" HTML-style comment fences to delimit injected content.
- Capability inventory: The skill has the capability to read and write files (via scripts) and execute local maintenance scripts in the ".claude/scripts/" directory.
- Sanitization: Parity and balance checks are performed post-sync using "verify-sync-adoption-parity.mjs".
- [DYNAMIC_EXECUTION]: The skill's maintenance workflow includes modifying the logic or configuration within the management scripts themselves before execution.
- Evidence: Step B2 requires editing ".claude/scripts/sync-hooks-to-skills.py" and ".claude/scripts/agent_protocol_matrix.py" to update synchronization tiers.
Audit Metadata