The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements an extension hook system that reads configurations from .specify/extensions.yml. It identifies commands to run based on YAML keys and instructs the agent to automatically execute mandatory hooks using the EXECUTE_COMMAND pattern. This allows for the execution of arbitrary skill commands defined in the project's data files.
[COMMAND_EXECUTION]: The execution flow begins by running a local PowerShell script located at .specify/scripts/powershell/check-prerequisites.ps1. While this is a local file, it represents a dependency on project-supplied executable code.
[INDIRECT_PROMPT_INJECTION]: The skill treats several project files (spec.md, plan.md, tasks.md, and constitution.md) as the 'sole source of intent' for its logic. Malicious content within these files could influence the agent's behavior during the assessment phase.
Ingestion points: Reads spec.md, plan.md, tasks.md, .specify/memory/constitution.md, and .specify/extensions.yml.
Boundary markers: None. The skill does not use delimiters or instructions to ignore potential injections within these artifacts.
Capability inventory: The skill can execute shell commands, run scripts, read files, and append content to the filesystem.
Sanitization: There is no logical sanitization of the input documents, though basic shell escaping is mentioned for user arguments.

speckit-converge