planify
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from several potentially untrusted sources, including repository configuration files (like
package.jsonandpyproject.toml), project documentation (CLAUDE.md), user-provided arguments, and external report annotations. This information is used to drive the planning logic and sub-agent exploration. - Ingestion points: SKILL.md (Phase 1 exploration of repo files, Phase 2 user interrogation, and processing of external report blocks).
- Boundary markers: The workflow features a mandatory
<HARD-GATE>in Phase 3, which requires explicit user approval of the summarized plan before any tools are invoked to write files or delegate tasks. - Capability inventory: The skill utilizes
Read,Glob,Grep,WebSearch,WebFetch,Agent, andSkilltools. - Sanitization: The skill relies on human-in-the-loop validation to identify and reject any malicious instructions that might be interpolated from the ingested data into the final plan.
Audit Metadata