launch-registry
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data including user-provided evidence and submission proposals, which could serve as vectors for embedded malicious instructions.
- Ingestion points: Processes 'pasted platform text', 'submission/status proposals', and 'early-access graduation evidence' during registry operations as described in the Data Sources and Instructions sections of SKILL.md.
- Boundary markers: The instructions explicitly label 'pasted platform text' as untrusted evidence, though no specific structural delimiters are enforced for strict isolation.
- Capability inventory: The skill possesses the capability to execute a local Python script (registry-events.py) and shell commands (git rev-parse) as part of its operational procedure.
- Sanitization: Requires all requests to be 'schema-valid' and mandates verification of the registry script, event schema, and system catalog before execution.
- [COMMAND_EXECUTION]: The skill executes local commands and scripts to manage the launch registry environment and data persistence.
- Evidence: Uses git rev-parse --show-toplevel for path resolution and environment configuration and executes registry-events.py to append events to the NDJSON data stream.
Audit Metadata