The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze untrusted session logs (referred to as "traces") which could contain malicious instructions designed to influence the auditor or its sub-agents.
Ingestion points: The skill reads session logs and trace files in Step 1, 3, and 3b using external Python scripts.
Boundary markers: The instructions for spawning sub-agents (Step 4) do not specify the use of clear delimiters or "ignore embedded instructions" headers when passing trace content to the axis agents.
Capability inventory: The skill utilizes shell execution via bash for indexing (Step 3), running security tools like gitleaks (Step 5.5), executing tests with pytest, and installing/modifying skills with npx skills (Step 6). It also performs file system writes to the ~/skills/wiki/ directory.
Sanitization: The skill includes an explicit sanitization step (Step 5.5) to replace internal IPs, domains, and credentials with placeholders before merging data into the persistent wiki.
[COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands and local Python scripts.
Evidence: Use of python3 ~/.pi/agent/skills/skill-call-extract/scripts/find-skill-sessions.py for indexing and session retrieval.
Evidence: Execution of gitleaks, uv run pytest, and npx skills add during the remediation and verification phase in Step 6.
[DATA_EXPOSURE]: The skill processes session logs that may contain sensitive user data or credentials.
Evidence: Step 5.5 acknowledges this risk by requiring the use of gitleaks and manual sanitization of internal project names and credentials before committing findings to a repository.

skill-call-diagnose