speckit-tasks
Fail
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill automatically executes a local bash script located at
.specify/scripts/bash/setup-tasks.shduring the initialization phase. - [REMOTE_CODE_EXECUTION]: The skill implements a 'hooks' system that reads
.specify/extensions.ymland instructs the agent to execute arbitrary commands defined in that file viaEXECUTE_COMMAND. This allows any repository to define commands that the agent will run automatically (mandatory hooks) or upon suggestion (optional hooks), posing a significant supply chain risk. - [PROMPT_INJECTION]: Indirect prompt injection vulnerability surface detected.
- Ingestion points: Reads
.specify/extensions.yml,plan.md,spec.md,data-model.md, andresearch.mdfrom the local workspace. - Boundary markers: No explicit delimitation or instructions to ignore embedded commands within the processed files are present.
- Capability inventory: The skill has the capability to execute shell scripts and arbitrary commands defined in the project configuration.
- Sanitization: No validation or sanitization is performed on the commands retrieved from the configuration files before execution.
Recommendations
- AI detected serious security threats
Audit Metadata