The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes complex bash and git commands to manage repository state across multiple worktrees.
Evidence includes a bash loop that traverses git worktrees using git -C and awk to identify unpushed branches.
The skill performs git push operations and handles branch logic (git symbolic-ref, git push origin HEAD:refs/heads/main) to ensure session work is synced to a remote repository.
[INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection as it processes, consolidates, and promotes data from session logs and corrections files.
Ingestion points: Reads from session-logs/*.md, corrections.md, and context.md during the wrap-up process.
Boundary markers: No explicit delimiters or boundary markers are defined to separate instruction logic from the data being processed in the logs.
Capability inventory: The skill possesses full file system write access within its scope and the ability to push commits to remote repositories.
Sanitization: There is no documented sanitization or validation of the content within the session logs before it is consolidated or promoted to 'durable pattern cards'.

wrapup