The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions mandate the automatic execution of commands retrieved from a local configuration file.
Evidence: "After emitting the block above you MUST actually invoke the hook and wait for it to finish before continuing. ... EXECUTE_COMMAND: {command}"
[DYNAMIC_EXECUTION]: The skill facilitates the runtime execution of dynamically sourced commands via its extension hook mechanism.
Evidence: The agent is instructed to read .specify/extensions.yml and execute the content of the command field for any mandatory hooks detected.
[INDIRECT_PROMPT_INJECTION]: The skill ingests multiple unvalidated project files, creating a significant attack surface for indirect prompt injection.
Ingestion points: .specify/extensions.yml, spec.md, plan.md, tasks.md, and /memory/constitution.md.
Boundary markers: There are no delimiters or instructions to prevent the agent from obeying commands embedded within these processed files.
Capability inventory: The skill possesses the ability to read files and execute shell commands through its hook execution logic.
Sanitization: No sanitization or safety checks are applied to the external content before it is used to influence agent behavior or command execution.

speckit-analyze