multica-projects-and-resources
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
multicaCLI tool to perform administrative actions on project entities, including lifecycle management (create, update, list) and resource attachment. - [DATA_EXPOSURE]: The skill manages sensitive configuration data, specifically absolute local filesystem paths and GitHub repository URLs, which are stored as project resources to be used in future agent tasks.
- [PROMPT_INJECTION]: The skill features a mechanism where project descriptions and resource metadata are injected into the agent's 'brief' and workdir configuration (
.multica/project/resources.json). This creates a surface for indirect prompt injection, as malicious instructions stored in these fields could influence the agent's behavior during task execution. - Ingestion points: Project descriptions and resource lists are fetched from the Multica server API/DB and written to local configuration files.
- Boundary markers: The skill mentions these are injected under specific headers like
## Project Context, providing some structural separation. - Capability inventory: The agent has access to the
multicaCLI viaBashfor subsequent operations. - Sanitization: No specific sanitization or validation of the injected content is mentioned in the provided instruction files.
Audit Metadata