kk-curate
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches the
kenkeeppackage from the public npm registry (npmjs.com) at runtime usingnpx kenkeep@latest. This is a well-known service for tool distribution. - [REMOTE_CODE_EXECUTION]: Executes a local JavaScript utility,
.ai/kenkeep/scripts/kk-detect-root.mjs, to resolve the project's root directory. - [COMMAND_EXECUTION]: Invokes several system utilities and tool-specific commands including
npx,uuidgen,mkdir, andrmfor session management and file cleanup. - [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests session logs from
.ai/kenkeep/_sessions/*.mdwhich contain external conversation data. - Ingestion points: Reads files matching
.ai/kenkeep/_sessions/*.mdto extract transcripts. - Boundary markers: Absent; there are no specific delimiters or instructions defined to prevent the AI from obeying instructions embedded within the logs.
- Capability inventory: The skill can perform file writes to the
nodes/directory, executenpxcommands, and remove files usingrm. - Sanitization: Employs
npx kenkeep@latest validateto ensure curator outputs conform to thecurator-outputJSON schema before the persistence phase.
Audit Metadata