The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands including ls, find, cat, and mkdir to discover project structure and read manifest files (e.g., package.json, Cargo.toml). These operations are restricted to the local environment and are necessary for the skill's stated purpose of mapping codebases.
[INDIRECT_PROMPT_INJECTION]: The skill processes project source code and directory structures which could contain malicious instructions designed to influence the agent. The risk is considered low as the skill's outputs are limited to generating a local markdown report and it does not have network access for exfiltration.
Ingestion points: Project files, directory names, and source code content encountered during the Discovery Phase (SKILL.md).
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded prompts in analyzed code.
Capability inventory: The skill possesses read access to the local file system and write access to a specific output path (./.gtd/CODEBASE.md).
Sanitization: There is no evidence of content sanitization or escaping when interpolating file content into the generated report.

codebase-overview