The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The instructions in reference/subagents/config-analyzer.md explicitly direct the subagent to locate and extract sensitive credentials. Specifically, it is tasked with identifying and documenting OPENAI_API_KEY, database passwords (*_PASSWORD), access tokens, and private keys found in .env and other configuration files. While intended for project analysis, recording these secrets in a plain-text markdown report (repo-analysis-YYYYMMDD.md) creates a high risk of accidental data exposure.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes large amounts of untrusted external data from local and remote repositories, creating a vulnerability to indirect prompt injection.\n
Ingestion points: The main agent and subagents read all files within a repository, including README.md, CLAUDE.md, and the entire source code of the target repository, as part of the DISCOVERY and PARALLEL ANALYSIS phases.\n
Boundary markers: There are no specified delimiters or instructions provided to help the agent distinguish between its analysis tasks and potentially malicious commands hidden in the code it is processing.\n
Capability inventory: The skill utilizes tools to clone repositories, read arbitrary local files, spawn multiple subagents via the Agent tool, and write aggregated findings to markdown files.\n
Sanitization: The skill does not implement any logic to filter or sanitize the content of the analyzed files before it is processed or included in the final output report.

repo-analyzer