The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements a hook system that reads instructions from .specify/extensions.yml. It parses this file and uses the EXECUTE_COMMAND directive to trigger shell commands or agent tools defined within that file. This allows project-level configuration to control the agent's execution flow dynamically.
[COMMAND_EXECUTION]: The skill automatically executes a local shell script located at .specify/scripts/bash/check-prerequisites.sh during the outline phase. Running scripts provided by the target repository can lead to unauthorized local code execution if the repository content is malicious.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect injection through project files.
Ingestion points: The agent reads .specify/extensions.yml and .specify/scripts/bash/check-prerequisites.sh from the repository root.
Boundary markers: None identified. The instructions do not specify sanitization or verification of the hook commands.
Capability inventory: The skill has the ability to execute shell commands (EXECUTE_COMMAND), run bash scripts, and create GitHub issues via MCP.
Sanitization: There is no evidence of sanitization for the commands extracted from the extensions file before they are passed to the agent for execution.

speckit-taskstoissues