Agent skill

Coral Extend

by Human-Agent-Society in Human-Agent-Society/CORAL

Add a new component to the CORAL framework itself — a new agent runtime under coral/agent/builtin/ (claudecode/codex/cursoragent style), a new CLI command in coral/cli/, a new bundled skill or…

Apache-2.0Auto-check passedAgent Workflows

Install Coral Extend

skills CLI
$ npx skills add Human-Agent-Society/CORAL --skill coral-extend -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install Human-Agent-Society/CORAL coral-extend --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/Human-Agent-Society/CORAL.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/coral-extend .claude/skills/coral-extend && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
coral-extend
GitHub stars
1.1k
Token cost
~1.6k tokens
SKILL.md length
695 words
Files
1
Skills in repo
7
Repo updated
First seen
Licence
Apache-2.0

At a glance

Add a new component to the CORAL framework itself — a new agent runtime under coral/agent/builtin/ (claudecode/codex/cursoragent style), a new CLI command in coral/cli/, a new bundled skill or…

  • Works in 5 steps: Create coral/agent/builtin/.py and… → Register the runtime in… → Decide the runtime's native shared-state… → …
  • Tasks that involve Subagents
  • SKILL.md covers Extending the grader…, A new agent runtime, A new CLI command and A new bundled skill or…, plus 3 more sections
  • Calls uv

What it does

Coral Extend is an agent skill from Human-Agent-Society/CORAL. Add a new component to the CORAL framework itself — a new agent runtime under coral/agent/builtin/ (claudecode/codex/cursoragent style), a new CLI command in coral/cli/, a new bundled skill or subagent template under coral/template/skills/ or coral/template/agents/, a new hook in coral/hooks/, a new field in coral/config.py, or a framework-level extension to the grader stack under coral/grader/. NOT for writing a per-task grader or adding an example task — use coral-new-task for that. NOT for debugging existing…

Its SKILL.md is about 1.6k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Agent Workflows, covering Subagents and Debugging. The repository describes itself as: Open-source autoresearch powered by autonomous coding agents. Run Claude Code, OpenCode, and Codex with grading, shared knowledge, and multi-agent evolution. Accepted at COLM 2026. The licence is Apache-2.0.

When your agent uses it

  • Tasks that involve Subagents
  • Tasks that involve Debugging

Example prompts

  • “/coral-extend”

Requirements

  • Python 3

Workflow steps

5 steps, taken from the first numbered list in SKILL.md.

  1. Create coral/agent/builtin/.py and subclass AgentRuntime (coral/agent/runtime.py). Existing runtimes are the canonical reference…
  2. Register the runtime in coral/agent/registry.py
  3. Decide the runtime's native shared-state directory name (.claude for Claude Code, .codex for Codex, etc.). The worktree symlink uses this…
  4. If the runtime needs special config plumbing (e.g. cursor_agent.json, opencode.json, gateway port), follow the opencode pattern: emit a…
  5. Add a smoke test in tests/test_.py modeled on tests/test_cursor_agent.py.

What it can do on your machine

Read from SKILL.md and the folder at commit 0123dfb. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • uv

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use uv, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Coral Extend loads about 1.6k tokens when it runs. Until then it costs about 143 tokens; SKILL.md has 695 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~143
When it runs · the whole SKILL.md, loaded when a task matches
~1.6k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from Human-Agent-Society/CORAL at commit 0123dfb, republished under its Apache-2.0 licence (© Human-Agent-Society). 695 words, ~1,593 tokens.

Download SKILL.mdSave it as .claude/skills/coral-extend/SKILL.md (or your agent's skills folder).
name
coral-extend
description
Add a new component to the CORAL framework itself — a new agent runtime under `coral/agent/builtin/` (claude_code/codex/cursor_agent style), a new CLI command in `coral/cli/`, a new bundled skill or subagent template under `coral/template/skills/` or `coral/template/agents/`, a new hook in `coral/hooks/`, a new field in `coral/config.py`, or a framework-level extension to the grader stack under `coral/grader/`. NOT for writing a per-task grader or adding an example task — use `coral-new-task` for that. NOT for debugging existing code — use `coral-debug`.

Extending the CORAL framework

For day-to-day debug / reproduce loops see the sibling coral-debug skill. For creating a new examples/<task>/ (seed + task.yaml + grader package) see coral-new-task. This skill covers adding new components to the CORAL package itself.

Extending the grader infrastructure

If you're writing a grader for a specific task, use coral-new-task. This section is only for changes to the grader framework under coral/grader/:

  • New helpers on TaskGrader (coral/grader/task_grader.py) — make sure they're useful to multiple existing example graders before adding.
  • New GraderInterface implementations (coral/grader/protocol.py / base.py) — the bar is high; the existing protocol covers everything we currently need.
  • Daemon-side changes (coral/grader/daemon.py) — concurrency, queue caps, worktree isolation, retry policy. Cover with tests/test_grader_daemon.py.
  • Built-in graders under coral/grader/builtin/ — function_grader.py is the only one today; not wired through task.yaml. New built-ins should justify why a TaskGrader subclass per task isn't enough.

A new agent runtime

Adding a new runtime (e.g. another coding-agent CLI) means three small files plus a registry entry.

  1. Create coral/agent/builtin/<name>.py and subclass AgentRuntime (coral/agent/runtime.py). Existing runtimes are the canonical reference — claude_code.py is the most complete; codex.py and cursor_agent.py are smaller and easier to mimic.
  2. Register the runtime in coral/agent/registry.py:
    python
    _RUNTIMES["my_runtime"] = MyRuntime
    _ALIASES["mine"] = "my_runtime"
    _DEFAULT_MODELS["my_runtime"] = "default-model-id"
  3. Decide the runtime's native shared-state directory name (.claude for Claude Code, .codex for Codex, etc.). The worktree symlink uses this; pass it through shared_dir so generate_coral_md(...) renders the right paths.
  4. If the runtime needs special config plumbing (e.g. cursor_agent.json, opencode.json, gateway port), follow the opencode pattern: emit a per-agent config file inside the worktree at startup.
  5. Add a smoke test in tests/test_<runtime>.py modeled on tests/test_cursor_agent.py.

Reference recent additions: PR #79 (cursor_agent), commit f6f266e (codex web_search config fix).

A new CLI command

CLI is an old-school argparse single-file dispatcher.

  1. Add a parser block in coral/cli/__init__.py::main(). Match the existing style — _HelpOnErrorParser, an epilog with Examples:, _CommandHelpFormatter. Add the new command name to _VISIBLE_COMMANDS so "did you mean?" suggestions work.
  2. Implement cmd_<name>(args: argparse.Namespace) -> None in the most-fitting module under coral/cli/:
    • start.py — agent lifecycle (start/resume/stop/status)
    • query.py — read-only inspection (log/show/notes/skills/runs)
    • eval.py — agent-side commands that mutate the worktree (eval/wait/diff/revert/checkout)
    • heartbeat.py — heartbeat configuration
    • ui.py — dashboard
    • author.py — init / validate Create a new module if none of those fit; keep imports lazy so coral --help stays fast.
  3. Wire the function into the commands = {...} dict at the bottom of main().
  4. If your command operates on a specific run, accept --task / --run via _add_run_args(parser) and resolve with coral.cli._helpers.find_coral_dir.
  5. Add an example to CLAUDE.md's Commands section.
Show full SKILL.md (295 more words)Show less

A new bundled skill or subagent template

These ship inside the package and are seeded into every run's .coral/public/skills/ (or agents/) by coral/workspace/project.py.

  • Skill — create coral/template/skills/<name>/SKILL.md with frontmatter name and description. Include scripts/ and references/ subdirs as needed; existing examples are deep-research, organize-files, skill-creator.
  • Subagent — create coral/template/agents/<name>.md (single markdown file). Existing examples are deep-researcher and librarian.
  • Add a test in tests/test_template.py if the rendering pulls in new template variables.

The seed copy is one-shot per run (if not dst.exists()), so iterating on template content during development means deleting <run_dir>/.coral/public/skills/<name>/ and re-running coral start, or just editing the destination directly for that run.

A new hook

Right now there's only coral/hooks/post_commit.py. If you add another hook:

  • Define a clear single entrypoint function (model on submit_eval).
  • Make it pure-function over coral_dir + agent_id where possible.
  • Atomic writes to .coral/public/ only; never write to a worktree from a hook.
  • Add coverage to tests/test_hooks.py.

Configuration changes

coral/config.py is dataclass-based and merged via OmegaConf. When adding a new field:

  1. Add it to the right dataclass (AgentConfig, GraderConfig, ...) with a sensible default.
  2. If it deserves runtime validation, add it to the __post_init__ of that dataclass.
  3. Cover the new field in tests/test_config.py.
  4. Update examples/<task>/task.yaml only if the field is task-author facing — internal knobs should stay defaulted.
  5. Mention it in CLAUDE.md if it changes user-visible behavior; otherwise leave the docs alone (CLAUDE.md describes invariants, not every flag).

Don't forget

  • Lint + test before pushing: uv run ruff check . && uv run ruff format . && uv run pytest tests/ -v.
  • Backward compatibility for run dirs. People resume old runs. Anything that reads from .coral/public/ must tolerate missing files (return defaults), not crash.
  • No agent-side git. All commits go through coral eval → submit_eval. Don't add helpers that shell out to git from agent context.

© Human-Agent-Society, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in .claude/skills/coral-extend of Human-Agent-Society/CORAL.

Open the folder on GitHubat commit 0123dfb

Compare with similar skills

Coral Extend next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Coral Extend compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Coral Extend this skillHuman-Agent-Society/CORAL1.1k—~1.6kAutomated safety check: PassApache-2.0
Astra Orchestratordonvito/codex-astra-luna-orchestrator1.7k—~3kAutomated safety check: PassApache-2.0
Adk Debuggoogle/adk-python22k—~837Automated safety check: PassApache-2.0
Astra Orchestratordonvito/codex-astra-luna-orchestrator1.7k—~3.1kAutomated safety check: PassApache-2.0
Skill Architectcuriositech/some_claude_skills244—~5.7kAutomated safety check: NotesMIT
Skill Architectcuriositech/some_claude_skills244—~6kAutomated safety check: NotesMIT

Similar skills

  • Astra Orchestrator

    donvito/codex-astra-luna-orchestrator

    Orchestrate complex Codex coding work for the Plus profile with GPT-6 Luna at max reasoning as planner/integrator, Luna subagents for exploration, implementation, testing, and research, and an Astra…

    1.7k GitHub stars~3k tokensUpdated 9 days ago
    Agent WorkflowsAuto-check passed
  • Adk Debug

    google/adk-python

    Official

    Diagnoses misbehaving ADK agents by inspecting sessions, events, tool calls, and the exact request that reached the model.

    22k GitHub stars~837 tokensUpdated today
    Agent WorkflowsAuto-check passed
  • Astra Orchestrator

    donvito/codex-astra-luna-orchestrator

    Orchestrate complex Codex coding work for the Pro profile with GPT-6 Astra at medium reasoning as planner/integrator, Luna subagents for exploration, implementation, testing, and research, and an…

    1.7k GitHub stars~3.1k tokensUpdated 9 days ago
    Agent WorkflowsAuto-check passed
  • Skill Architect

    curiositech/some_claude_skills

    Design, create, audit, and improve Claude Agent Skills with expert-level progressive disclosure.

    244 GitHub stars~5.7k tokensUpdated 1 mo ago
    Agent WorkflowsAuto-check: notes
  • Skill Architect

    curiositech/some_claude_skills

    Design, create, audit, and improve Claude Agent Skills with expert-level progressive disclosure.

    244 GitHub stars~6k tokensUpdated 1 mo ago
    Agent WorkflowsAuto-check: notes
  • Analyze Trajectory

    yologdev/yoyo-evolve

    Diagnoses a recurring failure such as a stuck task, repeated CI error or frequent reverts by sending sub-agents through the logs and returning one root-cause diagnosis.

    1.9k GitHub stars~3.6k tokensUpdated today
    Agent WorkflowsAuto-check passed

More from Human-Agent-Society/CORAL

  • Creating A Coral Task

    Human-Agent-Society/CORAL

    Author a new CORAL task — the three pieces that must line up (task.yaml, seed/, a packaged grader/), the coral init → coral validate → smoke-test loop, and how to pick a grader pattern (stdout…

    1.1k GitHub stars~2.2k tokensUpdated 1 mo ago
    Auto-check passed
  • Running Coral Experiments

    Human-Agent-Society/CORAL

    Run and manage CORAL experiments from the operator side — launch agents with coral start (dotlist overrides, model/count, tmux vs local), monitor with coral status / coral log / coral show / the web…

    1.1k GitHub stars~1.3k tokensUpdated 1 mo ago
    Auto-check passed
  • Coral Debug

    Human-Agent-Society/CORAL

    Verify and debug changes to CORAL itself — smallest reproduce loop per area (grader / daemon / CLI / hooks / manager / workspace / hub / template / config / web), where to look when something breaks…

    1.1k GitHub stars~1.8k tokensUpdated 1 mo ago
    Auto-check passed
  • Coral New Task

    Human-Agent-Society/CORAL

    End-to-end recipe for adding a new task under examples/ — the three pieces that have to line up (task.yaml, seed/, and grader/), what to put in each, the TaskGrader API surface, the coral validate →…

    1.1k GitHub stars~3.4k tokensUpdated 1 mo ago
    Auto-check passed
  • Promoting Dev To Main

    Human-Agent-Society/CORAL

    A skill your agent uses when preparing, reviewing, resolving conflicts for, or merging a CORAL release pull request from the long-lived dev branch into main.

    1.1k GitHub stars~588 tokensUpdated 1 mo ago
    Auto-check passed
  • Setting Up Coral

    Human-Agent-Society/CORAL

    One-time machine setup after installing the coral CLI — register local agent runtimes as named bindings with coral setup / coral setup agent, validate them with coral agents doctor (incl.

    1.1k GitHub stars~1.7k tokensUpdated 1 mo ago
    Auto-check passed

Questions about Coral Extend

What does Coral Extend do?

Add a new component to the CORAL framework itself — a new agent runtime under coral/agent/builtin/ (claudecode/codex/cursoragent style), a new CLI command in coral/cli/, a new bundled skill or…. Coral Extend is an agent skill from Human-Agent-Society/CORAL.py, or a framework-level extension to the grader stack under coral/grader/.

When should I use Coral Extend?

Coral Extend fits situations like: tasks that involve Subagents; tasks that involve Debugging.

How do I install Coral Extend in Claude Code?

Run `npx skills add Human-Agent-Society/CORAL --skill coral-extend -a claude-code`. Or copy the skill folder (.claude/skills/coral-extend in Human-Agent-Society/CORAL) into .claude/skills/coral-extend in your project. Claude Code loads it when a task matches its description.

How do I install Coral Extend in Codex?

Run `npx skills add Human-Agent-Society/CORAL --skill coral-extend -a codex`. Or copy the skill folder (.claude/skills/coral-extend in Human-Agent-Society/CORAL) into .agents/skills/coral-extend in your project. Codex loads it when a task matches its description.

Can I use Coral Extend in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Human-Agent-Society/CORAL --skill coral-extend -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/coral-extend, .gemini/skills/coral-extend, .github/skills/coral-extend and .opencode/skills/coral-extend in your project.

What does Coral Extend need to run?

Going by SKILL.md and its folder, Coral Extend needs the command-line tools its instructions call (uv). Our summary lists: Python 3.

Does Coral Extend access the network?

SKILL.md contains no URLs. Its commands use uv, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Coral Extend safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Coral Extend use?

Coral Extend is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Coral Extend use?

About 1.6k tokens (SKILL.md is roughly 6.4k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Coral Extend?

Skills that share tags, products or a category with Coral Extend: Astra Orchestrator (donvito/codex-astra-luna-orchestrator, 1.7k stars), Adk Debug (google/adk-python, 22k stars), Astra Orchestrator (donvito/codex-astra-luna-orchestrator, 1.7k stars) and Skill Architect (curiositech/some_claude_skills, 244 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Coral Extend?

Human-Agent-Society (a GitHub organization) maintains it in Human-Agent-Society/CORAL, which has 1,060 GitHub stars. The repository holds 7 skills in this directory. The repository was last updated on September 8, 2026.

Source: Human-Agent-Society/CORAL on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.