complete-everything
Warn
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: MEDIUMINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONPERSISTENCE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill relies on searching project-specific files (plans, completion reports) and conversation history to determine the scope of work. This creates an attack surface where malicious instructions embedded in those files could influence agent behavior.
- Ingestion points: The skill reads from
.cursor/plans/, conversation logs, and existing completion reports using grep-like tools (rg). - Boundary markers: There are no instructions to use delimiters or ignore instructions found within the ingested data.
- Capability inventory: The skill possesses the ability to execute shell commands (for testing and verification) and write to the filesystem (
.cursor/complete-everything-state.md). - Sanitization: No sanitization or validation steps are defined for the data retrieved from external project files.
- [COMMAND_EXECUTION]: The instructions direct the agent to "discover exact verification commands" from arbitrary files within the repository, including
README.md,Makefiles,AGENTS.md, andCLAUDE.md, and to execute these discovered commands during the verification phase. - [PERSISTENCE]: The skill references a mechanism that stores counters and state in the user's home directory at
~/.cache/kenji/completion-gate/. This allows the skill's state and behavior to persist across different project directories and execution sessions.
Audit Metadata