The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands to discover the repository structure and automation entry points. Phase 1 involves running commands like tree --gitignore, find, just --list, and make help to gather project context. The instructions also suggest that the agent should install software via a package manager if the tree utility is missing or outdated, which involves executing installation commands that may exceed the defined tool scope.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes arbitrary content from the local repository (e.g., README.md, documentation files) to generate its output. Ingestion points: Phase 1, Step 2 and 3 instruct the agent to read various repository files to build context. Boundary markers: The skill does not define specific delimiters or instructions to ignore potential commands or malicious content embedded in the files being read. Capability inventory: The agent possesses tools for file modification and limited shell execution, which could be targeted by instructions found in the ingested data. Sanitization: No evidence of sanitization or content validation for the data retrieved from the repository files is mentioned.

agentsmd-generator