The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a local bash command to check the modification timestamps of previous run artifacts to enforce an execution cadence limit.
Evidence: Bash snippet in SKILL_detail.md uses find and date to locate recently modified files in the tracks/_meta/ directory.
[INDIRECT_PROMPT_INJECTION]: As a verification pipeline, the skill is designed to ingest and analyze untrusted third-party skill files and project directories.
Ingestion points: Defined in Step 0, the skill reads SKILL.md files and arbitrary directory paths provided in the scope.
Boundary markers: The conductor relies on the constituent tools (e.g., steel-quench) to handle the parsing of these files.
Capability inventory: The skill has access to Bash, Read, Write, and Agent tools for orchestration.
Sanitization: The instructions focus on structured verdict aggregation (PASS/FAIL) rather than the sanitization of the untrusted inputs themselves.

pipeline-conductor