The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is configured to read and incorporate instructions from project files such as CLAUDE.md, AGENTS.md, and README.md. These files are potential vectors for indirect prompt injection if they contain malicious instructions designed to override the agent's behavior.
Ingestion points: SKILL.md directs the agent to locate and read CLAUDE.md, AGENTS.md, README.md, and other local project guides before performing work.
Boundary markers: The skill implements a procedural gate requiring the agent to state its execution plan to the user before modifying any files, providing an opportunity for user oversight.
Capability inventory: The skill possesses the capability to modify files, execute git commands, run build tools (pnpm), and perform framework-specific operations (node ace).
Sanitization: There is no evidence of specific sanitization or filtering logic applied to the content read from the project documentation files.
[COMMAND_EXECUTION]: The skill uses local shell commands to manage the development environment and repository state.
Evidence: SKILL.md contains instructions to use git status, rg --files, pnpm test, and node ace serve for project orchestration and verification.

maestro