speckit-tasks
Warn
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill is instructed to run a local shell script located at
.specify/scripts/bash/check-prerequisites.sh. Since this script resides within the project directory being analyzed, it may contain malicious code that would execute without verification. - [INDIRECT_PROMPT_INJECTION]: The skill processes the
.specify/extensions.ymlfile and is directed to execute any commands found under thehookskeys. This allows an attacker to influence the agent's behavior by placing malicious commands in the project's configuration. - Ingestion points:
.specify/extensions.yml(read during both pre-execution and post-generation phases). - Boundary markers: None. The skill does not implement delimiters or instructions to ignore embedded commands within the configuration data.
- Capability inventory: The skill uses the
EXECUTE_COMMANDdirective and dynamically generated slash commands to trigger agent tool calls. - Sanitization: None. The instructions explicitly state not to evaluate hook conditions, executing them based solely on their enabled status in the untrusted YAML file.
- [DYNAMIC_EXECUTION]: The skill dynamically constructs and executes commands at runtime based on external configuration. It transforms user-defined hook names from the YAML file into executable slash commands (e.g., converting dots to hyphens) and triggers them via the
EXECUTE_COMMANDmarker.
Audit Metadata