hw-run-policy

Name: hw-run-policy
Availability: InStock
Author: headwater-ai

Provides standing rulings and environment constraints for Headwater build-order run stages.

Updated Aug 8, 2026

Search Tags:#workflow #orchestration #cargo #github cli #adjudication #build-order #run policy

Ready to run

npx skills add https://github.com/headwater-ai/headwater --skill hw-run-policy-headwater-ai

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: hw-run-policy
Source: https://github.com/headwater-ai/headwater/tree/main/.claude/skills/hw-run-policy
Command: npx skills add https://github.com/headwater-ai/headwater --skill hw-run-policy-headwater-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Multi-stage build-order runs stall when each stage stops to re-derive shared rulings or rediscovers environment pitfalls, wasting turns and repeating costly mistakes. ## Core Features & Use Cases - Standing Rulings: Cites adjudication, refusal, milestone, and coordination decisions by identifier so a ruling tuned later moves in one place. - Environment Traps: Documents known failures with gh CLI, worktree isolation, cargo builds, and gating commands so no stage loses an hour relearning them. - Cost Discipline: Lists rules for blocking waits, fixed report blocks, and minimal tool output to control token and turn costs across a run. - Use Case: Before dispatching an hw-queue, hw-adjudicate, hw-build, hw-verify, or hw-integrate stage, load this policy so the stage follows the same adjudication and cost rules as every other stage. ## Quick Start Load the hw-run-policy skill before dispatching any stage of a Headwater build-order run.

Frequently Asked Questions about hw-run-policy

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I run a multi-stage build-order workflow without stages stalling?▼

Load the run policy before dispatching any stage so every stage shares the same standing rulings on adjudication, refusal, and coordination. This prevents stages from stopping to ask questions the rulings already answer.

What is the correct way to build the Headwater CLI during a run?▼

Build with cargo using the dev-release profile via the tools/hw-cargo wrapper, never the release profile, since release LTO is single-threaded and slow. The wrapper pools target directories across stages and caps concurrent compilers.

Why does gh issue view fail with a projectCards error?▼

The gh CLI hits a GraphQL projectCards deprecation on issue view and pr edit. Use gh api repos/headwater-ai/headwater/issues/<N> --jq .body to read and gh api -X PATCH to update bodies instead.

Should I run clippy locally before pushing a Headwater change?▼

No. Clippy is CI's gate, not a stage's; it compiles the workspace under a second toolchain and shares no cache with the run's builds. Push and read the CI result instead.

Why did my gate command report the wrong exit status?▼

Piping a gate into a filter reports the filter's exit status, not the gate's. Redirect output to a file and read the tail in a second call, keeping stdout and stderr in separate files.