agentleFS
Sign inStart free
metacircu1armetacircu1ar/hive_skills/skills/hive-parallel-review-loop/SKILL.md

hive-parallel-review-loop

Skill4 starsChanged 2 months agoMITView on GitHub

Run multiple independent repository implementation/review loops concurrently as hive_workflow processes backed by named hive_registry implementors and parallel reviewer panels. Use when a YAML manifest supplies disjoint repositories and shared plus per-assignment context, and no Codex subagents should be launched.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run N independent Hive implementation/review workflows across N disjoint repositories.

    line 12
  2. Very specific:

    Never launch Codex subagents.

    line 18
  3. Distinctive:

    Read the complete sibling hive-review-loop skill and its configuration contract before executing this skill.

    line 22
  4. Distinctive:

    If two repository runs are active and each has three reviewers, up to six reviewer provider turns can overlap after their implementor turns.

    line 67
  5. Distinctive:

    No repository, Hive run directory, or launcher-artifact directory may overlap another repository.

    line 71

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Hive Parallel Review Loop

Run N independent Hive implementation/review workflows across N disjoint repositories. Each workflow has one named registry implementor and a configurable panel of named registry reviewers. Hive runs the reviewers in each panel concurrently; this wrapper additionally runs multiple repository workflows concurrently up to max_parallel_runs.

Never launch Codex subagents. The only worker processes are hive_workflow commands, and only hive_registry may instantiate provider agents and own their sessions.

Read the complete sibling hive-review-loop skill and its configuration contract before executing this skill.

Execute

  1. Require unique assignment IDs, pairwise-disjoint absolute Git worktree roots, an external artifact root, a positive concurrency limit, defaults for registry agents, and complete shared plus per-assignment context.
  2. Confirm that every task authorizes the configured registry implementor to modify its repository.

  3. Validate the entire batch before launching any Hive run:

    python3 "<hive-review-loop-skill-dir>/scripts/run_hive_loops.py" parallel "<config.yaml>" --validate-only

  4. Run the batch:

    python3 "<hive-review-loop-skill-dir>/scripts/run_hive_loops.py" parallel "<config.yaml>"
  5. Report the aggregate Markdown summary, JSON summary, and every assignment's Hive run, reviewer panel, round reports, launcher logs, and terminal status.

The launcher completes independent assignments even when another assignment fails. Exit 0 requires every assignment to reach reviewer consensus and complete requested cleanup. Exit 1 means at least one assignment needs attention; its Hive run and sessions remain available. Exit 2 means batch preflight failed before any workflow launch.

Resume after fixing external failures or extending assignment round limits:

python3 "<hive-review-loop-skill-dir>/scripts/run_hive_loops.py" parallel "<config.yaml>" --resume

Existing assignments with run.json resume; assignments without a run start normally. Never delete or replace preserved run directories to manufacture a clean batch result.

Concurrency model

max_parallel_runs limits concurrent repository workflows, not reviewer count. If two repository runs are active and each has three reviewers, up to six reviewer provider turns can overlap after their implementor turns. Choose the limit from provider quotas and machine capacity.

No repository, Hive run directory, or launcher-artifact directory may overlap another repository. Each assignment receives its own Hive run and registry sessions; no agent conversation is shared across assignments.

Example activations

  • "Run this three-repository Hive manifest with two workflows at a time."
  • "Use the default reviewer panel everywhere, but override it for the billing repository."
  • "Resume only the unfinished Hive runs and produce a new aggregate summary."

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free