cahier-des-charges
Pass
Audited by Gen Agent Trust Hub on Mar 1, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE]: The skill consists entirely of Markdown files and configuration. There are no scripts (Python, JavaScript, Shell) or binaries included in the package, eliminating risks associated with malicious code execution.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from external specification files (
requirements.mdanddesign.md). While these files represent an untrusted data surface, the skill lacks any capabilities (such as tool use, network access, or file system modification) that could be exploited by injected instructions. - Ingestion points: Processes
.agent/specs/{feature-name}/requirements.mdanddesign.mdas defined in the workflow ofSKILL.md. - Boundary markers: Absent; the agent is instructed to read these files directly without specific delimitation.
- Capability inventory: None. The skill only generates text output based on the provided template.
- Sanitization: Absent; the skill relies on the underlying LLM's standard safety filters.
Audit Metadata