registry / @rschlaefli/manifest-mcp / 0.4.0

@rschlaefli/manifest-mcp@0.4.0

Manifest — a calm, shared AI-agent Kanban board. MCP server + singleton local daemon; agents share one live board across tools, and humans steer it with comments.

AI Security Review

scanned 4h ago · by lpm-firewall-ai

LPM treats this as warn-only first-party agent extension lifecycle risk. An explicit `manifest init` onboards a project by adding MCP/Claude configuration, a lifecycle hook, and bundled agent assets. The hook updates a local board and may persist approved plans; no unconsented install-time mutation or non-local endpoint was found.

Static reason

No blocking static signals were detected.

Trigger

User runs `manifest init`, then configured Claude lifecycle events occur.

Impact

Can modify the current project's MCP/Claude configuration and create board/plan files.

Mechanism

Explicit first-party Claude extension setup and local daemon orchestration.

Rationale

Source inspection confirms explicit, first-party agent configuration mutation and a persistent local daemon, but no concrete malicious behavior. Per policy this is a warn-level extension lifecycle risk, not a publish block.

Evidence

package.jsondist/index.mjs.agents/hooks/manifest-hook.mjsREADME.md.mcp.json.claude/settings.json.claude/skills.claude/commands/handoff.md.claude/commands/review.md.agents/plans/<id>.md~/.manifest/board.ndjson~/.manifest/daemon.json

Network endpoints2

127.0.0.1:<port>/api/health/api/*

@rschlaefli/manifest-mcp@0.4.0

AI Security Review

Decision evidence

Source & flagged code

Source & flagged code

Findings

Source & flagged code