The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill is designed to read sensitive local configuration files, including .env and other environment files, to identify and validate service dependencies.
Evidence: The instructions in SKILL.md (Step 1 and 2) explicitly direct the agent to locate the 'canonical local environment file' and trace 'environment lookups' from runtime code and provider clients.
Context: This access is necessary for the skill's primary purpose of environment validation.
Sanitization: The skill contains specific directives (Step 5) to 'strip URL userinfo, paths, query strings, headers, response bodies, credential values, and exception messages from output' and to 'keep [secret values] in memory.'
[COMMAND_EXECUTION]: The skill implements and executes a custom shell-based command within the repository context.
Evidence: The agent is instructed to add a 'doctor' recipe to a Justfile and execute it (Step 5 and 7) to verify the environment.
[DYNAMIC_EXECUTION]: The skill involves generating new executable logic (the doctor command and associated tests) which is then executed by the agent.
Evidence: Step 5 describes implementing the command using the repository's native language and Step 6 involves adding new tests for this implementation.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted repository files (docs, config, .env) which are used to generate the validation code.
Ingestion points: Justfile, .env, repository documentation, and setup scripts.
Boundary markers: No specific delimiters or 'ignore' instructions are defined for the content read from these files.
Capability inventory: The agent has filesystem write access (modifying Justfile, adding tests) and command execution capabilities.
Sanitization: The skill mandates strict redaction of sensitive data discovered during execution to minimize exposure risks.

add-just-doctor