speckit-implement
Warn
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local shell scripts and commands based on project configuration.
- Runs
.specify/scripts/bash/check-prerequisites.shfrom the repository root. - Executes arbitrary shell commands defined under the
hooks.before_implementandhooks.after_implementkeys in.specify/extensions.yml. - [REMOTE_CODE_EXECUTION]: The 'Extension Hooks' mechanism allows the repository's configuration to trigger arbitrary command execution. If a user clones an untrusted repository, the author of that repository can execute commands on the user's system via the mandatory hook (
optional: false) execution path. - [INDIRECT_PROMPT_INJECTION]: The skill processes multiple external files which dictate the agent's logic and command execution flow without sufficient sanitization or boundary markers.
- Ingestion points: Processes
.specify/extensions.yml,tasks.md,plan.md, and various checklist files. - Boundary markers: None are specified; the agent is instructed to directly 'Load and analyze' the context and 'Execute implementation following the task plan'.
- Capability inventory: Includes shell script execution, file creation/modification (ignore files), and git operations.
- Sanitization: No validation or sanitization of the commands or instructions contained within the ingested files is performed before execution.
Audit Metadata