Skip to content
Agentmart
Submit a tool

hm-l2-phase-execution

Execute GSD-style phase plans with wave-based parallelization and checkpoint recovery.

Skill by shynlee04

IndexedMIT★ 51
Write a review

Chalk Labs

Need help deploying hm-l2-phase-execution?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/shynlee04-hm-l2-phase-execution

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Execute GSD-style phase plans with wave-based parallelization and checkpoint recovery. Use when running a multi-plan phase, managing plan dependencies, recovering from mid-phase interruptions, when plans need to run in parallel, when execution needs checkpoint gates, or when phase work spans multiple sessions. Even when the user says "run the phase" or "execute the plan." Triggers: "execute phase", "run phase", "phase execution", "wave-based execution", "parallel plans", "checkpoint recovery", "plan dependencies". NOT for single-task execution or unstructured work.

Use cases: productivity

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found
Runs as
Shell scripts

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/hm-l2-phase-execution

What's inside

Skill name
hm-l2-phase-execution
Description
Execute GSD-style phase plans with wave-based parallelization and checkpoint recovery. Use when running a multi-plan phase, managing plan dependencies, recovering from mid-phase interruptions, when plans need to run in parallel, when execution needs checkpoint gates, or when phase work spans multiple sessions. Even when the user says "run the phase" or "execute the plan." Triggers: "execute phase", "run phase", "phase execution", "wave-based execution", "parallel plans", "checkpoint recovery", "plan dependencies". NOT for single-task execution or unstructured work.
License
MIT
Metadata
role: orchestrator · layer: 1 · pattern: P2 · version: 1.0.0
Instructions
167 lines in SKILL.md · view on GitHub

Before installing

  • Ships executable scripts

Agent Skills spec issues

  • allowed-tools: must be a space-separated string

Files (7)

  • SKILL.md
  • evals.json · evals/
  • rich-gate-scorecard.md · metrics/
  • checkpoint-recovery.md · references/
  • execution-state-template.md · references/
  • wave-protocol.md · references/
  • validate-skill.sh · scripts/

SKILL.md (first lines)

## Overview

Execute phase plans with wave-based parallelization and checkpoint recovery. Use when running multi-plan phases, managing plan dependencies, recovering from mid-phase interruptions, or coordinating parallel plan execution across sessions. Produces completed phase plans with per-task verification and atomic commits.

## The Iron Law

```
A phase is not a todo list. A phase is a directed acyclic graph of plans grouped into waves.
```

# Phase Execution
## On Load

1. Read `references/wave-protocol.md` — how to group plans into waves
2. Read `references/checkpoint-recovery.md` — mid-phase interruption recovery

## Phase Execution Protocol

## File-Backed Execution State Machine

Adapted from Nanostack conductor's claim/dependency/artifact/stale-lock pattern, but mapped to HiveMind/OpenCode project conventions. Use durable state files so a later agent can recover without trusting chat history.

| State artifact | Purpose | Notes |
|----------------|---------|-------|
| `.opencode/state/hivemind/phase-execution/<phase>/claims/<plan>.json` | Records active executor, started time, wave, dependencies | Atomic creation preferred; never overwrite another active claim without stale evidence |
| `.opencode/state/hivemind/phase-execution/<phase>/artifacts/<plan>.md` | Handoff summary, files changed, validations, blockers | Written before checkpoint or completion |
| `.opencode/state/hivemind/phase-execution/<phase>/done/<plan>.json` | Completion marker with verification evidence | Created only after plan verification passes |
| `.opencode/state/hivemind/phase-execution/<phase>/failures/<plan>.json` | Failure/blocker evidence | Blocks downstream waves until resolved |

Do not import `.nanostack/` paths. If the end-user project has a different state root, use it and document the adapter path in the artifact.

### Step 1: Load Plan Index

```bash
# Read phase plan index
ls .planning/phases/${PHASE}/
# Identify all PLAN.md files and their wave numbers
```

### Step 2: Validate Dependencies

```bash
# Check for circular dependencies
# Verify all depends_on references exist
# Confirm wave numbers are monotonic
```

**Gate:** If validation fails, STOP. Fix plans before executing.

### Step 2.5: Claim Runnable Plans

Before dispatching a plan in a wave, create a claim record with executor identity, command/session id if available, dependency list, and stale-after timestamp. If claim creation fails because another live executor owns it, skip that plan and report the owner. If a claim is stale, record the stale evidence before taking over.

### Step 3: Execute Wave by Wave

```
Wave 1: Plans with no dependencies → parallel dispatch
  → Wait for all to complete
  → Validate results
Wave 2: Plans depending on Wave 1 → parallel dispatch
  → Wait for all to complete
  → Validate results
...
Wave N: Final plans → parallel dispatch
  → Wait for all to complete
  → Final validation
```

### Step 4: Handle Failures

| Scenario | Action |
|----------|--------|
| Plan in wave fails | Abort subsequent waves; log failure; return DONE_WITH_CONCERNS |
| Plan returns BLOCKED | Escalate to orchestrator; do not proceed |
| Plan returns NEEDS_CONTEXT | Provide context; re-dispatch same plan |
| All plans succeed | Proceed to verification |

Every failure writes `failures/<plan>.json` with the failed command, output path, affected dependencies, and next recovery action. Never let a downstream wave consume a failed or missing upstream artifact.

### Step 5: Verify and Commit

```bash
# Run phase verification script if available
bash .planning/phases/${PHASE}/verify.sh 2>/dev/null

# Commit all changes with phase-scoped message
git add <phase-files>
git commit -m "phase: ${PHASE} — <summary>"
```

## Checkpoint Recovery

If interrupted mid-phase:

```bash
# Read phase state
.cat .planning/phases/${PHASE}/STATE.md 2>/dev/null

# Identify incomplete plans
# Re-execute only incomplete plans
# Do NOT re-execute completed plans
```

Recovery order:

1. Trust `done/*.json` markers with verification evidence.
2. Inspect `failures/*.json` and return blockers before executing downstream waves.
3. Inspect active claims; only resume/take over if the stale-after time has passed or the coordinator explicitly authorizes recovery.
4. Reconstruct missing artifacts from commits/files only when state markers are absent, and label reconstruction as degraded evidence.

## RICH Gate Source Decisions

| Source | Decision | Local adaptation |
|--------|----------|------------------|
| `garagon/nanostack` conductor | ADAPT | File-backed claims, dependency checks, artifacts, and stale-lock recovery mapped to `.opencode/state/hivemind/`. |
| `github/awesome-copilot` create-implementation-plan | ADAPT | Deterministic plan identifiers and measurable validation criteria are reinforced in wave artifacts. |
| GitHub agent skill resource model | ADAPT | State template and evals are bundled resources; SKILL.md remains the execution index. |

## Independence Notes

From shynlee04/hivemind-plugin, MIT.

More from shynlee04/hivemind-plugin

Signals

Stars
51
Forks
16
Open issues
21
Last commit (this folder)
May 7, 2026 (5 months ago)

From GitHub (fetched Oct 24, 2026).

Details

Version
1.0.0
License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from shynlee04

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get hm-l2-phase-execution running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying hm-l2-phase-execution?

CallGet help
Cite this page

hm-l2-phase-execution skill. Agent Mart. https://agentmart.dev/skills/shynlee04-hm-l2-phase-execution (last updated 2026-10-24)