ingest-inbox-item
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest external content from the
inbox/directory into a long-term corpus. - Ingestion points: Reads files from
inbox/(markdown, JSON, JSONL, transcripts). - Boundary markers: None explicitly defined in the workflow to isolate the
[body]when it is being processed by sub-skills liketag-by-topicorscore-intuition-density. - Capability inventory: The workflow involves
mvoperations (file system write/delete) and writing new files tocorpus/seeds/viaStep 6andStep 9inSKILL.md. - Sanitization: No evidence of input sanitization or escaping is mentioned in the workflow. The instructions explicitly state
[body: preserved verbatim from normalize-format output], meaning any malicious instructions within the source file could potentially influence the agent's behavior during the tagging, scoring, or deduplication phases.
Audit Metadata