devlog-to-linkedin-post
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEDATA_EXFILTRATIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill attempts to read configuration from
~/.liftli/voice.mdand.liftli/voice.mdto establish the user's audience and writing style preferences. - [EXTERNAL_DOWNLOADS]: The skill provides instructions for the user to install additional packages using
npx skills add liftli-ai/linkedin-agent-skillsand recommends adding a Model Context Protocol (MCP) server from the vendor's domain athttps://mcp.liftli.ai/mcp. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill ingests untrusted data from
git log --onelineoutput, pull request titles, and the.liftli/voice.mdfile within the repository (SKILL.md). - Boundary markers: Absent; the instructions do not specify delimiters to separate the ingested git material from the core prompt instructions.
- Capability inventory: The skill uses
git logto retrieve project history and has the ability to read local filesystem paths. - Sanitization: No explicit sanitization, filtering, or escaping is performed on the ingested git metadata before it is processed by the model.
- Ingestion points: The skill ingests untrusted data from
Audit Metadata