build-by-story
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from
tasks.mdandrequirements.mdto drive agent behavior, partition work, and schedule tasks. - Ingestion points: Reads
tasks.mdandrequirements.mdto perform deterministic derivation of task units (referenced inSKILL.mdandstory-unit-mode.md). - Boundary markers: The skill records a content hash of 'Global Constraints' at setup to prevent unauthorized plan modification and enforces an 'Iron Law' human-in-the-loop gate before proceeding between units.
- Capability inventory: Performs file writes to the
.skills/directory andtasks.md, executes shell commands viagit, and orchestrates sub-agents for implementation and review. - Sanitization: Uses structured parsing for specific identifiers (e.g.,
**CODE-N.M**) and topological sorting rather than interpreting freeform natural language for logic derivation.
Audit Metadata