speckit-taskstoissues
Fail
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a bash script located at the relative path
.specify/scripts/bash/check-prerequisites.sh. This allows any repository containing this file structure to execute arbitrary code on the user's system when the skill is invoked. - [DYNAMIC_EXECUTION]: The skill implements an extension hook system that reads
.specify/extensions.ymland automatically triggers commands using theEXECUTE_COMMANDdirective. This pattern allows external project configuration to control the agent's internal command execution without sufficient user review. - [INDIRECT_PROMPT_INJECTION]: The skill generates prompts and execution instructions using data directly ingested from the
.specify/extensions.ymlfile. - Ingestion points:
.specify/extensions.yml(hooks, descriptions, and prompts). - Boundary markers: Absent; content from the YAML is presented as authoritative instructions to the agent.
- Capability inventory: Shell script execution (bash), Git operations, and GitHub issue creation via MCP server.
- Sanitization: Absent; hook command names and prompt text are used without validation or escaping.
- [REMOTE_CODE_EXECUTION]: By executing scripts and triggering commands defined within the project's own files, the skill facilitates a vector for remote code execution if the agent is operating on a repository cloned from an untrusted source.
Recommendations
- AI detected serious security threats
Audit Metadata