The Agent Skills Directory

[SAFE]: The skill serves as a read-only source of truth (a 'contract') for board operations, providing checklists and definitions for other skills in the system to follow.
[COMMAND_EXECUTION]: The instructions specify that the agent should use established local automation scripts (such as scripts/claim-card.sh) and the GitHub CLI (gh) to interact with the board and repository. These actions are transparently documented as the primary mechanism for board management.
[PROMPT_INJECTION]: The skill processes user-controlled data in the form of Markdown card bodies. It explicitly defines mitigation logic for potential indirect prompt injection through 'filler detection' rules and the INVEST validation checklist. Evidence chain: (1) Ingestion occurs when reading card bodies from the board; (2) Boundary markers are established via the card's Markdown section structure; (3) Capabilities are limited to standard Git/CLI operations; (4) Sanitization is performed via explicit validation rules defined in the card body schema reference.

board-canon