The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructions direct the agent to read sensitive local files, specifically .env, .env.example, and cloud provider profiles (AWS, GCP, Azure). This access is necessary for its purpose of checking project configuration. The risk is mitigated by an explicit instruction to never report secret values, only their presence and the names of the keys.
[COMMAND_EXECUTION]: The agent uses shell utilities like grep, comm, and sort to perform a precise comparison between example and active environment files. It also checks Makefile and package.json for defined project targets to verify the local stack is ready to run.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from untrusted repository files which could contain malicious instructions designed to influence the agent.
Ingestion points: .env, .env.example, README.md, CLAUDE.md, Makefile, and package.json (SKILL.md).
Boundary markers: The skill does not define specific markdown delimiters for the read file content, but it does instruct the agent on how to process the data.
Capability inventory: Filesystem access and shell command execution.
Sanitization: The skill employs a strict regular expression (^[A-Z][A-Z0-9_]+) during key extraction from env files. This acts as a primary safety filter to ensure only valid environment variable keys are handled, preventing the agent from interpreting arbitrary text values inside the files as instructions.

check