The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external, potentially untrusted data from project repositories to generate documentation.
Ingestion points: The workflow in SKILL.md instructs the agent to inspect existing READMEs, manifests, public APIs, examples, and tests.
Boundary markers: The instructions do not specify any delimiters or safety warnings to ignore instructions embedded within the ingested files.
Capability inventory: The skill explicitly grants the agent the capability to execute code in SKILL.md (Workflow Step 5: "Run the smallest safe example or relevant check when feasible").
Sanitization: There is no mention of sanitizing or validating the content of the repository files before inspection or execution.
[COMMAND_EXECUTION]: The skill workflow includes an instruction to execute commands found within the project repository to verify examples. While the instruction specifies "safe" examples, this relies on the agent's judgment when interacting with potentially malicious code provided in a repository.

create-readme