The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a shell script find-polluter.sh designed to automate the identification of tests that create unwanted state (pollution) by iteratively running npm test on files matching a provided pattern. This is a standard utility for debugging flaky test suites.
[COMMAND_EXECUTION]: The SKILL.md file contains instructions and examples for diagnostic instrumentation in multi-layer systems. This includes shell commands like env | grep to verify environment variable propagation and macOS security tools to inspect keychain identities during codesigning failures.
[INDIRECT_PROMPT_INJECTION]: The debugging process involves the agent ingesting and analyzing external data such as error logs, stack traces, and system outputs which could theoretically contain malicious instructions.
Ingestion points: Phase 1 of the debugging process in SKILL.md, where the agent is instructed to read error messages and gather evidence from multi-component systems.
Boundary markers: Absent; the skill does not define specific delimiters for separating logs or error messages from the instruction context.
Capability inventory: The agent has the ability to execute shell commands (as shown in the find-polluter.sh script and instrumentation examples) and read project files.
Sanitization: Absent; the instructions focus on direct analysis of the raw error data for root cause identification.

systematic-debugging