The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions in SKILL.md (Phase 1, Step 4) encourage the agent to inspect environment variables and system keychains to verify secret propagation during debugging. Specifically, it suggests commands like env | grep IDENTITY and security list-keychains. While these are common developer debugging steps, they instruct the agent to access and potentially output sensitive configuration metadata and identity information.
[COMMAND_EXECUTION]: The utility script find-polluter.sh performs automated execution of npm test on files discovered via the find command. This involves a capability to execute arbitrary code defined within test files or project-level lifecycle scripts (e.g., pretest) based on user-provided file patterns.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze potentially untrusted external data such as error logs, stack traces, and component output.
Ingestion points: Phase 1 of the debugging process in SKILL.md explicitly instructs the agent to read error messages and logs from multi-component systems.
Boundary markers: The instructions do not specify the use of delimiters or provide warnings to ignore instructions that might be embedded within the logs or error messages.
Capability inventory: The agent has the capability to read local files and execute shell commands (e.g., via the provided bisection script and instrumentation examples).
Sanitization: There is no evidence of sanitization or escaping protocols for external content before it is processed by the agent.

tool-systematic-debugging