The Agent Skills Directory

[PROMPT_INJECTION] (LOW): The skill uses authoritative directives like 'The Iron Law' and 'ALWAYS'/'NEVER' to override default agent behavior and enforce a specific methodology.
[COMMAND_EXECUTION] (LOW): The file find-polluter.sh executes npm test on local files matching a user-provided pattern, facilitating arbitrary local code execution.
[CREDENTIALS_UNSAFE] (LOW): Documentation in SKILL.md includes example commands for inspecting macOS keychains (security list-keychains) and code-signing identities (security find-identity). While intended for diagnostic purposes, these templates instruct the agent on accessing sensitive system credentials.
[PROMPT_INJECTION] (LOW): The skill is susceptible to Indirect Prompt Injection (Category 8).
Ingestion points: System error messages, stack traces, and component logs (defined in SKILL.md Phase 1).
Boundary markers: Absent; the skill does not define delimiters for external data.
Capability inventory: Local code execution (npm test via script) and system state inspection (security commands, env access).
Sanitization: Absent; the agent is instructed to read error messages 'carefully' without specific filtering rules.

systematic-debugging