stitch-delivery-harness
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the execution of local Python scripts, such as
stitch_harness.pyandsetup_harness_runtime.py, to manage the design delivery lifecycle and perform visual comparisons. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from project specification files to determine the scope and parameters of its actions.
- Ingestion points: The skill reads project specifications from
.stitch/specs/<page-id>.jsonfiles. - Boundary markers: Explicit user approval gates are implemented at critical decision points (
AWAITING_ART_DECISION,AWAITING_USER_APPROVAL), and the skill requires strict matching of user responses against pre-defined values. - Capability inventory: The skill is capable of executing local scripts with arguments, writing evidence receipts, and performing file system operations related to project archiving.
- Sanitization: The workflow includes mandatory text normalization (NFKC) and strict enum-based validation for user-supplied decision strings to prevent command or logic injection.
Audit Metadata