plan-milestones
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from several local sources, including
vision.md,specs/INDEX.md, and milestone assessment files. While this creates a potential surface for indirect prompt injection, the risk is effectively mitigated by comprehensive structural validation rules (S1–S8 and A1–A7) and mandatory human approval gates (HARD-GATE) before any roadmap update or milestone closure is finalized. - [DATA_EXFILTRATION]: The skill facilitates integration with Notion for milestone sign-offs. It reads configuration from
docs/agents/issue-tracker.mdto identify the target Notion workspace. This behavior is consistent with the skill's documented purpose of program management and does not involve accessing sensitive system credentials or unauthorized network communication. - [COMMAND_EXECUTION]: The instructions reference internal commands such as
/refresh-roadmap-statusand/assess-milestone. These appear to be companion agent skills or built-in commands used to orchestrate complex planning workflows. There is no evidence of arbitrary shell command execution or unsafe interpolation of user input into system calls.
Audit Metadata