The Agent Skills Directory

[DATA_EXPOSURE]: The skill reads .specweave/increments/*/metadata.json and tasks.md files to identify increments ready for closure. This access is restricted to the local project environment and does not involve sensitive user credentials or system paths.
[COMMAND_EXECUTION]: The script uses standard shell utilities including find, sort, jq, and grep to parse local data. No privileged commands (like sudo) or network-facing tools (like curl) are used.
[PROMPT_INJECTION]: There is a minimal indirect prompt injection surface as the skill processes local data files.
Ingestion points: metadata.json and tasks.md within the .specweave/increments directory.
Boundary markers: None present.
Capability inventory: Shell commands for file discovery and data parsing; delegation to the sw:done skill.
Sanitization: The use of jq -r for structured extraction and grep -c for counting patterns prevents data from being executed as arbitrary shell commands.

close-all