A workflow system for coordinating several coding agents on one larger task. A lead agent makes decisions while Hydra dispatches workers, tracks dependencies, retries work, checks health, and collects structured results.

Breaking a task into dependent or parallel assignments, managing agent worktrees, retrying workers, collecting results, and merging completed work.

It organizes parallel work and makes worker outcomes predictable, including when an agent is stuck or needs more information.

hydra — Skill for Claude Code and Codex by blueberrycongee: A workflow system for coordinating several coding agents on…

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods

npx agentmods add skills/blueberrycongee/termcanvas/hydra

Any agent

npx skills add blueberrycongee/termcanvas --skill hydra

Clone the repo

git clone --depth 1 https://github.com/blueberrycongee/termcanvas

Made for: Claude Code, Codex.

Or install skills, the plugin that ships this one along with the rest of its 7 skills, 1 hook.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/blueberrycongee/termcanvas/hydra.svg)](https://agentmods.dev/skills/blueberrycongee/termcanvas/hydra)

Your own site

<a href="https://agentmods.dev/skills/blueberrycongee/termcanvas/hydra"><img src="https://agentmods.dev/badge/skills/blueberrycongee/termcanvas/hydra.svg" alt="Measured on agentmods" height="20"></a>

Per session 33 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 2,075 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5	$0.00033	$0.02075
Opus 5	$0.00016	$0.01038
Sonnet 5	$0.00007	$0.00415
Haiku 4.5	$0.00003	$0.00208

Measured 5d ago against content hash e04ba70ed22a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

hydra scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/skills/hydra/SKILL.md · 196 lines

How it starts

The opening of the file, as written. The whole thing — 196 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Hydra Orchestration Toolkit

Hydra is a Lead-driven orchestration toolkit. You (the Lead agent) make strategic decisions; Hydra handles operational management (dispatch, retry, health checks, result collection).

Sub-agents output semantic intent (done/needs_rework/replan), not routing information. Hydra manages the lifecycle; you decide what happens next.

Why this design (vs. other coding-agent products)

SWF decider pattern, specialized for LLM deciders. Hydra is the AWS SWF / Cadence / Temporal decider pattern. hydra watch is PollForDecisionTask; the Lead is the decider; lead_terminal_id enforces single-decider semantics.
Parallel-first, not bolted on. dispatch + depends_on + worktree + merge are first-class. Other products (Factory.ai's Droid, Amp, Claude Code subagents) treat parallelism as open research; Hydra makes it the default.
Typed result contract. Workers publish a schema-validated result.json (outcome: completed | stuck | error, optional stuck_reason: needs_clarification | needs_credentials | needs_context | blocked_technical). Other products return free-text final messages and require downstream parsing.
Lead intervention points. hydra reset --feedback lets the Lead actually intervene at decision points instead of being block-and-join. A stale or wrong run is one reset away.

Lead operational rules

Core rules:

Root cause first. Fix the real implementation problem before changing tests, fixtures, or mocks.
Do not hack tests to force a green result. If a test is wrong, fix it honestly.
No silent fallbacks or swallowed errors. Surface failure with outcome=stuck or outcome=error.

Agent launch rule:

When dispatching Claude/Codex through TermCanvas, start a fresh agent terminal with termcanvas terminal create --prompt "...".
Do not use termcanvas terminal input for task dispatch — it is not a supported automation path.

Telemetry polling:

Treat hydra watch as the main polling loop. Do not infer progress from terminal prose.
Before deciding wait / retry / takeover, query:
- termcanvas telemetry get --workflow <workflowId> --repo .
- termcanvas telemetry get --terminal <terminalId>
- termcanvas telemetry events --terminal <terminalId> --limit 20
Watch the derived telemetry states: awaiting_contract means the worker has not yet published result.json; stall_candidate means the worker may be hung. Trust derived_status and task_status over terminal prose.

Read the full file on GitHub · 196 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

5d ago First seen · 196 lines · 33 tokens per session scan A e04ba70ed22a

Subscribe to this mod's changes

hydra is a skill published in the GitHub repository blueberrycongee/termcanvas (393 stars, last pushed 3mo ago), licensed MIT. It adds 33 tokens to every session and 2,075 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Building agents Planning agent-orchestration ai-agents canvas claude-code desktop-app developer-tools

ai-evaluation-patterns

Guides expert-level ai evaluation patterns implementation: ai-ml and testing decision frameworks, production-ready patterns, and concrete templates for ai evaluation patterns workflows. Use when the user asks about ai evaluation patterns, ai evaluation patterns configuration, or ai-ml best practices for ai projects.…

FerroxLabs/wayland · 86 tokens

automation-architect

Workflow automation design using Zapier, Make, and n8n covering trigger-action architecture, multi-step automation workflows, error handling strategies, personal versus business automation patterns, ROI calculation, and a library of common automation recipes. Use when the user asks about automation architect, related…

FerroxLabs/wayland · 90 tokens

sales-icp

Build a comprehensive Ideal Customer Profile (ICP) for any B2B business - firmographic, technographic, behavioral, pain-point, budget, and channel dimensions, plus negative ICP, 100-point scoring rubric, buyer personas, prospecting playbook, and a draft first-outreach message that inherits sales-outreach Phase 0…

FerroxLabs/wayland · 181 tokens

sales-qualify

Qualify a sales lead using BANT (Budget/Authority/Need/Timeline) and MEDDIC (Metrics/Economic Buyer/Decision Criteria/Decision Process/Identify Pain/Champion) frameworks against publicly available signals using OSINT only - no scraping of platforms whose ToS forbid it (LinkedIn §8.2, Glassdoor, G2, Capterra…

FerroxLabs/wayland · 186 tokens

fix-issues

Auto-fix GitHub issues labeled as bugs: fetch open bug issues, analyze feasibility, fix code, and submit PRs. One issue per invocation. Use when: (1) User says "/fix-issues", (2) User asks to fix GitHub issues.

FerroxLabs/wayland · 59 tokens

fix-sentry

Auto-fix high-frequency Sentry issues: fetch issues > N occurrences, analyze stack traces, fix code, create GitHub issues, and submit PRs. Supports user feedback issues (event.type "default") with attachment analysis (logs, screenshots) when includefeedback=true. Use when: (1) User says "/fix-sentry", (2) User asks to…

FerroxLabs/wayland · 83 tokens