The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several shell commands to gather project metadata. Specifically, it uses git branch -a, git remote show origin, and grep patterns to detect branch structures and repository state.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it reads and summarizes external, potentially untrusted project data to create high-level instructions (AGENTS.md) that influence future agent behavior.
Ingestion points: The agent is directed to read various project files including package.json, CHANGELOG.md, Cargo.toml, go.mod, and CI/CD configuration files (.github/workflows/).
Boundary markers: Absent. The instructions do not specify the use of delimiters or specific tags to isolate ingested file content from the agent's logical processing.
Capability inventory: The skill utilizes shell execution for analysis and has file-writing capabilities to create or update the project's AGENTS.md file.
Sanitization: Absent. There are no explicit instructions to escape, filter, or validate the content harvested from the project files before it is incorporated into the generated documentation.
[DATA_EXPOSURE]: The skill analyzes local configuration files and environment variable templates (e.g., .env.example) to identify necessary project parameters. While this is restricted to local metadata, it facilitates the aggregation of project-specific technical details into a single document.

agents-md-composer