The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides the find-polluter.sh shell script, which automates the process of identifying tests that leave behind unwanted side effects by sequentially executing npm test on files matching a specified pattern.
[DATA_EXPOSURE]: Documentation in SKILL.md and root-cause-tracing.md encourages gathering technical evidence by logging environment variables, working directories, and system identities (e.g., using macOS security tools). It specifically demonstrates a secure way to verify the presence of secrets using shell parameter expansion (${VAR:+SET}) to avoid printing the actual values.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data such as error messages, stack traces, and log files. While it lacks explicit boundary markers to delimit this untrusted content, the skill's capabilities (executing tests and Git commands) are restricted to the local development context.
Ingestion points: Error messages, log outputs, and source code files analyzed during Phase 1 in SKILL.md.
Boundary markers: Not explicitly defined in the instructions.
Capability inventory: Execution of npm test, git init, codesign, and shell utilities found in find-polluter.sh and SKILL.md.
Sanitization: No specific sanitization or filtering of external data is described.
[PROMPT_INJECTION]: The instructions use strong, authoritative language (e.g., "The Iron Law", "MUST complete each phase") to ensure the agent adheres to the debugging process. These instructions function as behavioral constraints rather than malicious overrides of safety guidelines.

systematic-debugging