The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by processing external data to drive its investigative steps. Ingestion points: Verbatim logs (SKILL.md Step 1), database query results (references/chain.md Step 4), and source code/commit history (SKILL.md Step 3). Boundary markers: The skill requires logs to be "quoted verbatim" but does not specify formal delimiters or instructions for the agent to ignore instructions embedded in the logs. Capability inventory: Access to git (worktree, bisect, log), psql, grep, and local script execution. Sanitization: No explicit instructions for sanitizing ingested data before interpolation into prompts.
[COMMAND_EXECUTION]: The workflow requires the use of several command-line tools for technical investigation. Evidence includes the use of git worktree add, git status --porcelain, git log, git bisect, grep -rn, and psql to verify assumptions and reconstruct timelines.
[DYNAMIC_EXECUTION]: Step 11 in SKILL.md requires the execution of a local Python script (prose-lint.py) bundled with the skill's assets. This is a local dynamic execution pattern used for quality control on the final report.

rca