The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a research detection mechanism (Phase 0.45) that automatically extracts findings from the conversation history to influence subsequent implementation phases.
Ingestion points: Conversation history is scanned to identify and extract data from previous deep-research sessions in file SKILL.md.
Boundary markers: The provided instructions do not specify any delimiters or safety markers to isolate extracted content from the skill's own instructions.
Capability inventory: The skill is granted Write, Edit, and Bash permissions, allowing it to modify source files and execute system commands in Phase 3 and Phase 6.
Sanitization: There are no evident sanitization or verification steps for the data retrieved from the chat history before it is used to guide file modifications.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform standard developer workflow operations.
Backup management: Uses rsync to create file backups and executes a local script ./scripts/verify-backup.sh to ensure integrity.
Git operations: Employs git commands such as git log, git add, and git tag for version control management.
File filtering: Uses grep to verify plugin status in the PLUGINS.md file.

plugin-improve