The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands including ls, git log, and grep to verify the state of the project and implementation. These operations are limited to information gathering within the local repository and do not include administrative or destructive commands.
[PROMPT_INJECTION]: The skill implements a workflow for indirect prompt injection (Category 8) by ingesting data from untrusted or external plan and execution files to update global agent instructions in CLAUDE.md.
Ingestion points: Reads plan.html, plan.md, execution.md, and implementation-notes.md from the project directory.
Boundary markers: The skill does not define specific delimiters or "ignore instructions" wrappers for the content promoted to CLAUDE.md.
Capability inventory: The skill has capabilities to read files, execute git logs, and write/update markdown files across the repository.
Sanitization: There is no technical sanitization or escaping of the ingested content; it relies on a manual "admission test" and human-in-the-loop approval before applying changes.

qstack-plan-close