Agentpathfinder

Unlimited tamper-evident cryptographic task tracking for AI agents. Green = proven complete. Red = failed or incomplete. Free forever — no usage caps, no tel...

This package appears internally consistent with its stated purpose (local, tamper‑evident task tracking). Before installing or running it, consider these practical checks: - Inspect the shipped code (you can review the included Python files) and confirm you trust the publisher/repo before running pf install or pip install. The package runs locally and writes secrets to your home directory. - Run it in an isolated environment (VM, container, or dedicated user account) if you plan to bind steps that run system commands (examples show docker and subprocess usage). Bound step functions can execute arbitrary commands — that's a feature for automation but also a risk if agents or step code are untrusted. - Protect the data directory (~/.agentpathfinder or ~/.agentpathfinder/pathfinder_data) with filesystem permissions and avoid using it on multi‑user/shared machines unless you accept that anyone with access to that directory can read shards. The free tier is explicitly "tamper‑evident, not tamper‑proof." If you need stronger guarantees, follow the project's stated upgrade path (hosted vault/TEE) when available. - Verify the repository/source you install from (signed releases or a trustworthy GitHub repo). There are small doc mismatches (data path and license string) worth noting but not security‑critical by themselves. - Consider running a test with harmless tasks first and inspect the files created under your home directory to confirm what is stored and ensure no unexpected network activity occurs. If you want additional assurance, provide the exact install command/source URL you intend to use (clawhub or pip/GitHub), and I can re-check for remote install URLs or publish artifacts that would change the assessment.

Type: OpenClaw Skill Name: agentpathfinder-agent-task-tracker-free Version: 1.2.7 The AgentPathfinder bundle is a legitimate cryptographic task-tracking utility for AI agents. It implements XOR-based secret sharing to shard a master key across multiple task steps and uses HMAC-SHA256 for tamper-evident audit logging. Analysis of the Python source code (agent_runtime.py, task_engine.py) and scripts confirms that all data operations are strictly local, confined to the ~/.agentpathfinder directory, with no evidence of network exfiltration, unauthorized file access, or malicious execution. The bundle includes a comprehensive SAFETY.md that transparently explains its cryptographic choices and addresses potential false positives from security scanners.