speckit-constitution

Warn

Audited by Gen Agent Trust Hub on Oct 12, 2026

Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill implements a hook mechanism that reads command names from a local configuration file (.specify/extensions.yml) and instructs the agent to run them using the EXECUTE_COMMAND directive. This pattern constitutes dynamic command invocation where the specific tools or actions triggered are determined by external file content at runtime.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from several potentially untrusted sources to update the project constitution and related templates without employing boundary markers or input sanitization.
  • Ingestion points: Content is read from .specify/extensions.yml, .specify/memory/constitution.md, multiple template files in .specify/templates/, README.md, and docs/quickstart.md.
  • Boundary markers: There are no delimiters or specific instructions provided to the agent to ignore embedded commands or instructions within the retrieved file content.
  • Capability inventory: The skill possesses the ability to read and write files within the project's configuration directory and can trigger other agent commands through its hook system.
  • Sanitization: The skill does not perform any validation, escaping, or filtering of the data processed from external files before using it to influence agent output or execution.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 12, 2026, 03:23 PM