The Agent Skills Directory

[COMMAND_EXECUTION]: The skill implements an extension hook system that reads commands from .specify/extensions.yml located in the project root. Mandatory hooks marked as optional: false are executed automatically via the EXECUTE_COMMAND: directive.
Evidence: The Pre-Execution Checks and Mandatory Post-Execution Hooks sections instruct the agent to parse the YAML file and invoke identified commands without user confirmation for mandatory entries.
[COMMAND_EXECUTION]: The skill executes a script placeholder {SCRIPT} from the repository root to initialize the environment and capture feature metadata.
Evidence: Step 1 of the Outline phase requires running {SCRIPT} with --json --paths-only arguments to determine FEATURE_DIR and FEATURE_SPEC paths.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from feature specifications and requirement checklists, which could contain instructions designed to influence the agent's behavior during the clarification or integration phases.
Ingestion points: FEATURE_SPEC file, /memory/constitution.md, and FEATURE_DIR/checklists/requirements.md (specifically GitHub task-list items).
Boundary markers: The skill does not define strict delimiters or "ignore instructions" markers when reading the feature spec for the ambiguity scan.
Capability inventory: The agent has the ability to execute shell commands (via hooks/scripts), write to the filesystem (updating the spec and checklist), and interactively prompt the user.
Sanitization: No specific sanitization or filtering is performed on the content of the specification files before they are parsed for ambiguity or updated.

speckit-clarify