join-project
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted project documentation to generate summaries, creating a potential surface for indirect prompt injection. * Ingestion points: Reads 'llmwiki/wiki/index.md', 'llmwiki/wiki/log.md', and various concept files within 'llmwiki/wiki/concepts/'. * Boundary markers: Absent; documentation content is summarized without the use of explicit delimiters or instructions to ignore embedded commands. * Capability inventory: Includes local file reading, directory listing ('ls'), content searching ('grep'), and the ability to append or write new markdown files to the wiki's draft and log sections. * Sanitization: Absent; the skill does not explicitly filter or validate the ingested markdown content for malicious instructions before processing.
- [COMMAND_EXECUTION]: The skill utilizes standard shell commands such as 'test', 'ls', and 'grep' to validate the project structure and search for internal documentation links. These commands are hardcoded and restricted to the local workspace and agent tool directories (~/.agents/skills/).
Audit Metadata