The Agent Skills Directory

[DATA_EXPOSURE]: The skill reads local project documentation including files in docs/strategy/, VISION.md, ROADMAP.md, and .claude/knowledge/experiments.md. This behavior is limited to the local filesystem and is the core intended purpose of the skill to provide a strategic overview.
[PROMPT_INJECTION]: The skill ingests untrusted data from local Markdown files (Indirect Prompt Injection surface). A malicious file within the project could theoretically attempt to influence the agent's summary output. However, since the skill has no network access (preventing exfiltration), no file-writing permissions (preventing persistence), and no code execution capabilities, the risk is minimal.
[COMMAND_EXECUTION]: The skill description mentions searching for files and checking modification times. These are standard file system operations performed by the agent to fulfill the request and do not involve executing dangerous shell commands or acquiring elevated privileges.

project-review