The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes conversation transcripts from local history files, which constitutes a surface for indirect prompt injection if those logs contain adversarial instructions.
Ingestion points: The scripts/collect_sessions.py script reads session history in JSONL format from ~/.claude/projects/ and ~/.codex/sessions/.
Boundary markers: While transcripts are condensed and redacted, the system does not use explicit delimiters to isolate ingested transcript text from the agent's internal instructions.
Capability inventory: The skill possesses capabilities to draft skill modifications and generate local HTML reports; however, applying modifications requires explicit user confirmation.
Sanitization: The skill implements a comprehensive secret redaction mechanism in scripts/collect_sessions.py and performs HTML escaping in scripts/render_report.py to prevent data leakage and XSS.
[COMMAND_EXECUTION]: The script scripts/collect_sessions.py executes a local system command.
Evidence: The skill uses subprocess.run to execute git rev-parse --show-toplevel for the purpose of identifying the current repository root during the inventory phase.

skill-doctor