coding-agent-orchestration-harness

Plugin coding-agent-orchestration-harness from marketplace agent-harness in ebigunso/agent-harness. Commands read from the repo's manifest — yesterday.

Skills (18)

• codex-harness-bootstrap

  Install or refresh Codex custom-agent TOML profiles and the minimal AGENTS.md loader block for the coding-agent orchestration harness.

• counsel

  Conduct and limits of a session opened as Counsel, the role that talks with the person directing the work about what the product does or should do and records what that person ratifies as the grounds for implementation. Use only in a session opened as Counsel, for value discussions, initiative brief

• durable-docs-authoring

  Authoring and restructuring rules for durable documentation and decision records (ADRs). Use when writing, revising, renaming, or restructuring a document meant to stay accurate over time, or when an ADR is being considered, proposed, or drafted, including when a plan, an older record, or an instruc

• engineering-quality-baselines

  Routes validation depth and evidence expectations, including required checks, through progressive disclosure across architecture gates, testing and validation, test authoring, review rubric, latent-risk routing, and long-horizon audit. Use when doing non-trivial implementation, PR/code review, or bu

• git-workflow

  Safe Git procedure for branch gates, coherent commits, and non-interactive, non-destructive defaults. Use when planning or executing commit-affecting work, deciding whether a commit may proceed or how to split changes, routing shared-state Git mutations through the Orchestrator, creating or updating

• improvement-loop

  Self-improvement loop for Orchestrator-driven workflows. Use when a hard gate was missed (required validation or evidence skipped, a wrong done/blocked state, a safety or policy gate miss), when a run audited against the owner's value documents made a wrong stop or skipped a decision that needed the

• orchestration-harness

  Load for coding-related tasks in repositories using the coding-agent orchestration harness. Source of truth for deciding direct handling vs full harness workflow, subagent delegation, validation, review, rule updates, skill governance, and final reporting.

• plan-format

  Standard execution plan format for decomposing non-trivial work into Task_X tasks with owns/depends_on, explicit validation ownership, and explicit Task Waves (parallel dispatch sets). Use when asked to make a plan, break down tasks, outline implementation steps, execute multi-step work, or when non

• playwright-cli

  Provider-specific execution path for browser automation with playwright-cli (commands, run-code, artifact conventions). Use when a browser automation provider has been selected for the task and it is playwright-cli.

• playwright-e2e-evidence

  Standard spec and evidence format for browser E2E and visual validation. Use when a task or plan requires UI or E2E acceptance evidence, such as user-flow verification, visual or responsive checks, or browser-based validation of a change.

• rulebook

  Maintains repository rule files under docs/coding-agent/rules/*.md. Use when bootstrapping, repairing, migrating, refreshing, updating repo rules, applying repo-local rule_candidates, adding repo reference documents, or staging harness migration candidates.

• runtime-adapter-contract

  Maintains runtime-specific adapter files for the harness. Use when editing Copilot, Claude, or Codex agent definitions, manifests, role names, tool permissions, prompt length, or bootstrap behavior.

• skills-maintenance

  Maintain first-party skills and govern skill updates. Use when creating, reviewing, or updating a skill's trigger description, structure, provenance handling, bundled references, or maintenance rules, and when deciding whether a change belongs in SKILL.md, references/*, or should instead go through

• subagent-report-contract

  The contract for the YAML report a Worker subagent returns as its final message, and the validator for it. Use when producing a Worker report, validating one, or defining or updating Worker reporting requirements.

• subagent-strategy

  Delegation and parallel-subagent strategy for Orchestrator. Use when you are deciding how to split work across Researcher/Worker/Reviewer/Auditor, writing or refining subagent prompts, running parallel research/review, consolidating multiple subagent outputs, or when subagent calls are drifting in s

• value-documents

  The forms of the value documents a target repository carries (product philosophy, engineering philosophy, initiative brief, discussion notes) and the companion beside each philosophy, what each means, where each lives, and who may write or change each. Use when drafting, changing, locating, handing

• wave-integration

  Orchestrator-owned integration checklist for Worker waves. Use after one or more Worker reports return, before Reviewer dispatch, or before final closeout.

• workspace-troubleshooting

  Triage runbook for workspace failures, with references for stale view or branch mismatch, GitHub CLI auth failures, and unexpected external file changes. Use when the user's view of the repository differs from yours, gh auth fails mid-task, files change without your edits, or a command fails or beha

Subagents (6)

• Auditor

  Read-only subagent for independent judgement of the Orchestrator's work, following the mandate named in a fixed location-only dispatch.

• Counsel

  Explicitly selected main-thread Counsel for the coding-agent orchestration harness. Holds the value-level discussion with the owner about what the product does or should do, at the level of behaviour and decisions, and writes the value documents. A separate session from the Orchestrator; never dispa

• Orchestrator

  Explicitly selected main-thread Orchestrator for the coding-agent orchestration harness. Plans non-trivial work, dispatches Researcher/Worker/Reviewer/Auditor agents, integrates results, requires validation/review evidence, routes git through git-workflow, routes skill governance through skills-main

• Researcher

  Research-only subagent. Gathers workspace context and returns findings to the dispatching session, either plan-fill inputs for the Orchestrator or a behaviour-and-decision-level report for Counsel. Modifies no existing file the dispatch does not grant, does not write plan files, and does not interac

• Reviewer

  Review-only subagent. Reviews changes after an implementation wave/phase completes, or a draft plan before approval. Verifies objectives/acceptance criteria, best practices, and validation evidence. May run bounded UI evidence checks using a selected browser automation provider such as playwright-cl

• Worker

  Execution subagent for a single atomic Task_X within owns scope. May run bounded Worker UI probes when assigned UI/frontend work; Reviewer owns independent acceptance evidence. Implements changes, runs required validation, returns a strict YAML report (subagent-report-contract), and does not perform

Similar plugins