The Agent Skills Directory

[DATA_EXPOSURE]: The skill accesses project-specific files such as spec.md, plan.md, and tasks.md within a user-defined FEATURE_DIR. This is consistent with its stated purpose of analyzing requirements.
[COMMAND_EXECUTION]: The instructions include a note on shell escaping for arguments (e.g., handling single quotes), which is a best practice for preventing command injection in environments where these arguments might be passed to a shell, rather than a malicious pattern.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from project files which could theoretically contain malicious instructions.
Ingestion points: Reads spec.md, plan.md, and tasks.md from the local filesystem.
Boundary markers: None explicitly defined for the ingested text.
Capability inventory: Limited to reading files and writing markdown checklists to the FEATURE_DIR/checklists/ directory.
Sanitization: No specific sanitization logic is mentioned for the ingested text, though the skill is instructed to summarize content rather than output it raw.

bb-checklist