The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local shell script .specify/scripts/bash/check-prerequisites.sh --json to gather project context during the setup phase.
[COMMAND_EXECUTION]: The skill implements an 'Extension Hooks' feature that identifies and executes commands defined in the project's .specify/extensions.yml file. This includes both 'Automatic' (mandatory) and 'Optional' hooks.
[REMOTE_CODE_EXECUTION]: By instructing the agent to execute arbitrary commands sourced from a local configuration file (extensions.yml), the skill creates a mechanism where a malicious project repository can run arbitrary code via the agent's shell access.
[PROMPT_INJECTION]: The skill processes untrusted input from $ARGUMENTS and various local project files (e.g., spec.md, plan.md) to derive checklist items and clarifying questions.
Ingestion points: Reads from $ARGUMENTS, .specify/extensions.yml, and local Markdown files in the feature directory.
Boundary markers: No explicit boundary markers or delimiters are used to wrap the untrusted content before processing.
Capability inventory: The skill has the ability to execute shell scripts, run dynamic commands via hooks, and write new files to the filesystem.
Sanitization: The skill provides minimal sanitization by suggesting the escape of single quotes in arguments.

speckit-checklist