The Agent Skills Directory

[COMMAND_EXECUTION]: Several agents, including the build-verifier, commit-helper, debugger, orchestrator, and test-runner, utilize the Bash and Task tools. These agents are instructed to execute standard development commands such as npm run build, git diff, and various testing or debugging scripts. This behavior is consistent with the skill's primary purpose as a developer toolbox and does not exhibit malicious intent.
[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection because agents like the code-reviewer and debugger ingest untrusted content from the local filesystem (e.g., source code, stack traces, and build logs).
Ingestion points: Files are read via Read, Glob, and Grep tools in code-reviewer.md, debugger.md, and test-runner.md.
Boundary markers: The instructions do not define explicit delimiters or 'ignore' instructions for the processed data.
Capability inventory: The skill has broad capabilities including file writes (Write, Edit), shell execution (Bash), and project coordination (Task).
Sanitization: There is no explicit sanitization or validation of the ingested code content.
Assessment: While these ingestion points exist, they are inherent to the functionality of development tools, and the risk is considered low in the context of the intended use case.

developer-toolbox