10 actions in 5 pieces of software · ranked by words, meaning and measured trust · 559 ms

1. 1claude-securityverified and signedpassport

   skillanthropics/claude-plugins-officialrecord

   • claude-security/agent/scan-researcher

     Sub-agent scan-researcher (in the claude-security plugin)

2. 2pack-agentfound in code, not yet checkedno passport yet

   MCP serversakikotgw/pack-agent

   • pack_list

     List every pack exported to .agent-pack/exports/ and every pack currently installed (via install-ledger) in this project -- like a modpack launcher's "my instances" screen. Use before pack_show/pack_install to see what already exists.

   • pack_install

     Install an existing .pack.json to all detected harnesses (skills + experience hooks).

   • pack_export

     Scan project and write .agent-pack/exports/<name>.pack.json (does not install).

   • packagent_project

     Project a pack into .agent-pack/modpacks and index the SQLite catalog. Does not dsh plugin add.

   • pack_diff

     Compare two pack.json or lock.json files.

   • pack_sync

     Full pipeline: scan -> portable pack -> install to detected harnesses. Or install-only with from=existing pack.

3. 3io.github.TheAiSingularity/agentic-researchfound in code, not yet checkedno passport yet
   • research

     Run a full research query through the local engine.

4. 4research-agentfound in code, not yet checkedno passport yet

   MCP serverjonathanmoregard/research-agent

   • retry_research

     Re-run the scanner on a previously quarantined report.

5. 5dossier-research-mcpverified and signedpassport

   MCP servergit+https://github.com/fledgeling-co/dossier-research-mcpnpmrecord3,658 downloads/month

   • agent_create

     Create a PERSISTED custom research agent (the Managed Agents API, the other Gemini research surface). Unlike a Deep Research run, this gives you a reusable agent with a Linux sandbox that can run code, write files, and carry your house methodology across every run. Constraint worth knowing: at previ