codebase-teach
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [DATA_EXPOSURE_&_EXFILTRATION]: The skill is configured to read from
~/.claude/skills/_templates/feature/_template.md. While this is described as a template path for the Claude execution harness, accessing files within a user's home directory or application configuration folders is a sensitive operation that accesses data outside the immediate project scope. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the project source tree to extract business rules and domain knowledge, which could lead to an indirect prompt injection if the analyzed code contains malicious instructions.
- Ingestion points: Analyzes bounded contexts from the project source tree and reads the
CLAUDE.mdproject instruction file. - Boundary markers: None explicitly defined in the provided instruction snippet to delimit external code from agent instructions.
- Capability inventory: The skill has the capability to write files to the project directory (
ai-context/features/<context>.md) and produce a report (teach-report.md). - Sanitization: No specific sanitization or filtering logic is mentioned for the ingested source code content before it is processed or written to new files.
Audit Metadata