The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from existing project documentation (README.md, CONTRIBUTING.md, and other docs) which could contain hidden instructions intended to influence agent behavior during the audit or rewrite process.
Ingestion points: SKILL.md directs the agent to "Read the current README top to bottom", grep through tests and CI files, and analyze existing documentation for failure modes.
Boundary markers: The skill does not define specific delimiters or instructions to treat project content strictly as data rather than potential commands.
Capability inventory: The agent is empowered to rewrite project files, modify test assertions, and execute local shell commands, including the project's test suite.
Sanitization: No instructions are provided for sanitizing or escaping content ingested from project files before it is processed by the agent.
[COMMAND_EXECUTION]: The skill relies on the execution of local shell commands to audit the project state and verify the impact of changes.
Evidence: Instructions in SKILL.md include executing git status, git diff, grep for strings in CI/test files, and running the project's test suite ("Run every test that reads the README or CONTRIBUTING").
[DYNAMIC_EXECUTION]: The skill encourages the agent to generate and run logic to verify the integrity of the documentation.
Evidence: SKILL.md suggests the agent use "a quick script" to verify that relative links and assets (src, srcset) in the README resolve correctly.

better-readme