Pilot Task Chain

Chain tasks into sequential pipelines across agents. Use this skill when: 1. You need multi-step workflows where each step depends on previous results 2. You...

This skill is instruction-only and generally coherent for chaining pilot-protocol tasks, but before installing you should: 1) Ensure pilotctl on PATH and the pilot daemon are from a trusted source (verify versions/signatures) and run pilotctl daemon start as directed. 2) Install jq (examples use it) and ensure bash is available. 3) Provide the agent name variables used in examples (e.g., FETCH_AGENT, TRANSFORM_AGENT, STORE_AGENT or AGENT_1) — these are not declared in the registry metadata. 4) Be mindful that intermediate results are passed between agents; do not send sensitive secrets or private data through these pipelines unless you trust all participating agents and the pilot-protocol environment. 5) Confirm you trust the pilot-protocol skill/homepage and the AGPL-3.0 license implications. If you need higher assurance, ask the publisher to update registry metadata to list jq and required env vars explicitly and to provide a reproducible provenance for pilotctl.

Type: OpenClaw Skill Name: pilot-task-chain Version: 1.0.0 The skill provides a framework for chaining agent tasks using a 'pilotctl' binary, but the documentation in SKILL.md promotes a significant shell injection vulnerability. The provided bash examples pass raw task results ($RESULT) directly into subsequent command-line arguments without sanitization or proper escaping. This pattern allows a potentially malicious or compromised agent in the pipeline to execute arbitrary shell commands on the host by returning a payload (e.g., using backticks or command substitution) as its task result.