next-task
Warn
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill implements a workflow that reads and executes a 'finish sequence' of shell commands defined in
docs/agents/git-workflow.md(Step 8.7). This behavior allows for the execution of arbitrary code defined within a project's documentation files. - [COMMAND_EXECUTION]: The skill performs file system modifications (
rm,rmdir) and version control operations (git commit,git add,git rm) based on the contents of the.tasks/directory and instructions in the skill body. - [INDIRECT_PROMPT_INJECTION]: The skill has a high surface area for indirect injection as it processes untrusted data from task files and workflow documentation.
- Ingestion points: Reads task files from
.tasks/NN-*.mdand workflow instructions fromdocs/agents/git-workflow.md(SKILL.md). - Boundary markers: None; the agent is instructed to read the full task file and execute implementing code or sequences directly.
- Capability inventory: Shell command execution, file deletion (
rm), directory removal (rmdir), and git operations (add,commit,rm). - Sanitization: None; the skill does not specify validation or escaping for the commands read from the workflow file.
Audit Metadata