The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructions specify reading configuration files matching the .env* pattern to generate documentation updates. This pattern includes sensitive files such as .env, .env.local, and .env.production, which often contain hardcoded secrets, API keys, or private credentials. Accessing these files creates a risk of exposing secret values in the resulting documentation or during the multi-model consensus process.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from git diffs, git logs, and source code files to generate documentation updates, creating a surface for indirect prompt injection.
Ingestion points: The skill ingests data from changed source files (identified via git diff), commit messages (git log), and existing markdown documentation.
Boundary markers: The instructions do not define any delimiters or system-level warnings to separate the untrusted repository content from the agent's instructions during the analysis phase.
Capability inventory: The skill utilizes the Edit tool for file modifications and the Bash tool for repository operations, which provides a path for malicious instructions in code to affect the file system.
Sanitization: No sanitization, validation, or escaping of the extracted code content is specified before it is processed for documentation generation.

sc-readme