Agent skill

CCS Task Delegation

by kaitranntt in kaitranntt/ccs

Hands simple, deterministic tasks such as typo fixes, tests and small refactors to cheaper models through the ccs CLI, choosing a profile from your config.

MITAuto-check passedAgent Workflows

Install CCS Task Delegation

skills CLI
$ npx skills add kaitranntt/ccs --skill ccs-delegation -a claude-code

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

GitHub CLI
$ gh skill install kaitranntt/ccs ccs-delegation --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/kaitranntt/ccs.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/ccs-delegation .claude/skills/ccs-delegation && 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
ccs-delegation
GitHub stars
2.9k
Token cost
~1.8k tokens
SKILL.md length
652 words
Files
3 (incl. references)
Skills in repo
1
Repo updated
First seen
Licence
MIT

At a glance

Hands simple, deterministic tasks such as typo fixes, tests and small refactors to cheaper models through the ccs CLI, choosing a profile from your config.

  • Works in 7 steps: Parse override flag → Discover profiles → Analyze task requirements → …
  • Delegating typo fixes, small refactors or test additions to a cheaper model
  • SKILL.md covers Core Concept, User Invocation Patterns, Agent Response Protocol and Decision Framework, plus 4 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

The skill routes simple, deterministic tasks to cheaper models through the CCS CLI instead of handling them directly. Say 'use ccs' followed by a task and the agent picks a profile from `~/.ccs/config.json`, enriches the prompt with file context, runs `ccs` with that profile and the task, and reports back. A follow-up can continue the same session with the `:continue` form, and `--glm` or `--kimi` forces a profile.

Profile choice follows keyword analysis of the task. Words about thinking, analysis or debugging point to a reasoning profile, words about architecture or whole codebases point to a long-context one (kimi for both), and typo, test, refactor, update and fix words point to a low-cost one, with glm as the fallback. The agent reads the config first and stops with a hint to run `ccs doctor` if CCS is not configured. The description excludes complex architecture, security-critical code, performance optimization and breaking changes. A template and a troubleshooting reference are bundled, and the excerpt ends mid-protocol.

When your agent uses it

  • Delegating typo fixes, small refactors or test additions to a cheaper model
  • Forcing a specific CCS profile for a task
  • Continuing a delegated session with a follow-up request

Example prompts

  • “Use ccs to fix the typos in docs/setup.md.”
  • “Use ccs --glm to add unit tests for the date helpers.”
  • “use ccs:continue to commit the changes”

Requirements

  • The CCS command-line tool, with profiles configured in ~/.ccs/config.json

Workflow steps

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

  1. Parse override flag
  2. Discover profiles
  3. Analyze task requirements
  4. Select profile
  5. Enhance prompt
  6. Execute delegation
  7. Report results

What it can do on your machine

Read from SKILL.md and the folder at commit f45fa92. 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

    No scripts in the folder and no shell commands in SKILL.md.

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

  • Network

    No URLs in SKILL.md.

    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

CCS Task Delegation loads about 1.8k tokens when it runs, and up to ~3.7k if it reads all its reference files. Until then it costs about 113 tokens; SKILL.md has 652 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~113
When it runs · the whole SKILL.md, loaded when a task matches
~1.8k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~3.7k

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 kaitranntt/ccs at commit f45fa92, republished under its MIT licence (© kaitranntt). 652 words, ~1,794 tokens.

Download SKILL.mdSave it as .claude/skills/ccs-delegation/SKILL.md (or your agent's skills folder). This skill also uses 2 other files; get the full folder from GitHub.
name
ccs-delegation
description
Auto-activate CCS CLI delegation for deterministic tasks. Parses user input, auto-selects optimal profile (glm/kimi/custom) from ~/.ccs/config.json, enhances prompts with context, executes via `ccs {profile} -p "task"` or `ccs {profile}:continue`, and reports results. Triggers on "use ccs [task]" patterns, typo/test/refactor keywords. Excludes complex architecture, security-critical code, performance optimization, breaking changes.
version
3.0.0

CCS Delegation

Delegate deterministic tasks to cost-optimized models via CCS CLI.

Core Concept

Execute tasks via alternative models using:

  • Initial delegation: ccs {profile} -p "task"
  • Session continuation: ccs {profile}:continue -p "follow-up"

Profile Selection:

  • Auto-select from ~/.ccs/config.json via task analysis
  • Profiles: glm (cost-optimized), kimi (long-context/reasoning), custom profiles
  • Override: --{profile} flag forces specific profile

User Invocation Patterns

Users trigger delegation naturally:

  • "use ccs [task]" - Auto-select best profile
  • "use ccs --glm [task]" - Force GLM profile
  • "use ccs --kimi [task]" - Force Kimi profile
  • "use ccs:continue [task]" - Continue last session

Examples:

  • "use ccs to fix typos in README.md"
  • "use ccs to analyze the entire architecture"
  • "use ccs --glm to add unit tests"
  • "use ccs:continue to commit the changes"

Agent Response Protocol

For /ccs [task]:

  1. Parse override flag

    • Scan task for pattern: --(\w+)
    • If match: profile = match[1], remove flag from task, skip to step 5
    • If no match: continue to step 2
  2. Discover profiles

    • Read ~/.ccs/config.json using Read tool
    • Extract Object.keys(config.profiles) → availableProfiles[]
    • If file missing → Error: "CCS not configured. Run: ccs doctor"
    • If empty → Error: "No profiles in config.json"
  3. Analyze task requirements

    • Scan task for keywords:
      • /(think|analyze|reason|debug|investigate|evaluate)/i → needsReasoning = true
      • /(architecture|entire|all files|codebase|analyze all)/i → needsLongContext = true
      • /(typo|test|refactor|update|fix)/i → preferCostOptimized = true
  4. Select profile

    • For each profile in availableProfiles: classify by name pattern (see Profile Characteristic Inference table)
    • If needsReasoning: filter profiles where reasoning=true → prefer kimi
    • Else if needsLongContext: filter profiles where context=long → prefer kimi
    • Else: filter profiles where cost=low → prefer glm
    • selectedProfile = filteredProfiles[0]
    • If filteredProfiles.length === 0: fallback to glm if exists, else first available
    • If no profiles: Error
  5. Enhance prompt

    • If task mentions files: gather context using Read tool
    • Add: file paths, current implementation, expected behavior, success criteria
    • Preserve slash commands at task start (e.g., /cook, /commit)
  6. Execute delegation

    • Run: ccs {selectedProfile} -p "$enhancedPrompt" via Bash tool
  7. Report results

    • Log: "Selected {profile} (reason: {reasoning/long-context/cost-optimized})"
    • Report: Cost (USD), Duration (sec), Session ID, Exit code

For /ccs:continue [follow-up]:

  1. Detect profile

    • Read ~/.ccs/delegation-sessions.json using Read tool
    • Find most recent session (latest timestamp)
    • Extract profile name from session data
    • If no sessions → Error: "No previous delegation. Use /ccs first"
  2. Parse override flag

    • Scan follow-up for pattern: --(\w+)
    • If match: profile = match[1], remove flag from follow-up, log profile switch
    • If no match: use detected profile from step 1
  3. Enhance prompt

    • Review previous work (check what was accomplished)
    • Add: previous context, incomplete tasks, validation criteria
    • Preserve slash commands at start
  4. Execute continuation

    • Run: ccs {profile}:continue -p "$enhancedPrompt" via Bash tool
  5. Report results

    • Report: Profile, Session #, Incremental cost, Total cost, Duration, Exit code
Show full SKILL.md (234 more words)Show less

Decision Framework

Delegate when:

  • Simple refactoring, tests, typos, documentation
  • Deterministic, well-defined scope
  • No discussion/decisions needed

Keep in main when:

  • Architecture/design decisions
  • Security-critical code
  • Complex debugging requiring investigation
  • Performance optimization
  • Breaking changes/migrations

Profile Selection Logic

Task Analysis Keywords (scan task string with regex):

PatternVariableExample
/(think|analyze|reason|debug|investigate|evaluate)/ineedsReasoning = true"think about caching"
/(architecture|entire|all files|codebase|analyze all)/ineedsLongContext = true"analyze all files"
/(typo|test|refactor|update|fix)/ipreferCostOptimized = true"fix typo in README"

Profile Characteristic Inference (classify by name pattern):

Profile PatternCostContextReasoning
/^glm/ilowstandardfalse
/^kimi/imediumlongtrue
/^claude/ihighstandardfalse
otherslowstandardfalse

Selection Algorithm (apply filters sequentially):

profiles = Object.keys(config.profiles)
classified = profiles.map(p => ({name: p, ...inferCharacteristics(p)}))

if (needsReasoning):
  filtered = classified.filter(p => p.reasoning === true).sort(['kimi'])
else if (needsLongContext):
  filtered = classified.filter(p => p.context === 'long').sort(['kimi'])
else:
  filtered = classified.filter(p => p.cost === 'low').sort(['glm', ...])

selected = filtered[0] || profiles.find(p => p === 'glm') || profiles[0]
if (!selected): throw Error("No profiles configured")

log("Selected {selected} (reason: {reasoning|long-context|cost-optimized})")

Override Logic:

  • Parse task for /--(\w+)/. If match: profile = match[1], remove from task, skip selection

Example Delegation Tasks

Good candidates:

  • "/ccs add unit tests for UserService using Jest" → Auto-selects: glm (simple task)
  • "/ccs analyze entire architecture in src/" → Auto-selects: kimi (long-context)
  • "/ccs think about the best database schema design" → Auto-selects: kimi (reasoning)
  • "/ccs --glm refactor parseConfig to use destructuring" → Forces: glm (override)

Bad candidates (keep in main):

  • "implement OAuth" (too complex, needs design)
  • "improve performance" (requires profiling)
  • "fix the bug" (needs investigation)

Execution

Commands:

  • /ccs "task" - Intelligent delegation (auto-select profile)
  • /ccs --{profile} "task" - Force specific profile
  • /ccs:continue "follow-up" - Continue last session (auto-detect profile)
  • /ccs:continue --{profile} "follow-up" - Continue with profile switch

Agent via Bash:

  • Auto: ccs {auto-selected} -p "task"
  • Continue: ccs {detected}:continue -p "follow-up"

References

Template: CLAUDE.md.template - Copy to user's CLAUDE.md for auto-delegation config Troubleshooting: references/troubleshooting.md

© kaitranntt, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 2 other files (references) in .claude/skills/ccs-delegation of kaitranntt/ccs.

  • SKILL.md
  • CLAUDE.md.template
  • references/troubleshooting.md

Open the folder on GitHubat commit f45fa92

Compare with similar skills

CCS Task Delegation 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.

CCS Task Delegation compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
CCS Task Delegation this skillkaitranntt/ccs2.9k—~1.8kAutomated safety check: PassMIT
Fable Foremanolsenbrands/fable-foreman143—~5.2kAutomated safety check: PassMIT
Frontier Model Handoffkerpopule/hermes-jev-skills1.1k—~1.5kAutomated safety check: WarnMIT
Antigravityyuting0624/antigravity-for-claude-code375—~9.1kAutomated safety check: PassMIT
Provider Integrationhex/claude-council857—~635Automated safety check: PassMIT
Openrouter Context Optimizationjeremylongshore/tons-of-skills-marketplace2.8k—~2.4kAutomated safety check: PassMIT

Similar skills

  • Fable Foreman

    olsenbrands/fable-foreman

    Turns the lead model into a foreman that plans, routes and verifies while cheaper Claude, Codex or Grok workers do the typing, using a per-machine routing card.

    143 GitHub stars~5.2k tokensUpdated 16 days ago
    Agent WorkflowsAuto-check passed
  • Frontier Model Handoff

    kerpopule/hermes-jev-skills

    Chooses which paid frontier model seat should take a task already judged hard, hands it off with proper context, and keeps a watch on the delegated run.

    1.1k GitHub stars~1.5k tokensUpdated yesterday
    Agent WorkflowsAuto-check: warnings
  • Antigravity

    yuting0624/antigravity-for-claude-code

    Run the Antigravity CLI (Gemini) as a collaborating AI inside Claude Code, with intelligent model routing across the software development lifecycle.

    375 GitHub stars~9.1k tokensUpdated 5 days ago
    AI & LLM EngineeringAuto-check passed
  • Provider Integration

    hex/claude-council

    Adds new AI providers to claude-council, configures provider API settings, troubleshoots provider connections, and documents the provider script interface.

    857 GitHub stars~635 tokensUpdated yesterday
    AI & LLM EngineeringAuto-check passed
  • Openrouter Context Optimization

    jeremylongshore/tons-of-skills-marketplace

    Optimize context window usage for OpenRouter models to reduce cost and improve quality.

    2.8k GitHub stars~2.4k tokensUpdated today
    AI & LLM EngineeringAuto-check passed
  • FreeRide Free Model Manager

    Shaivpidadi/FreeRide

    Configures OpenClaw to use free OpenRouter models, setting the best one as primary and adding ranked fallbacks so rate limits do not interrupt work.

    238 GitHub starsUsed in 2 repos~1.1k tokens
    AI & LLM EngineeringAuto-check passed

Works with

Questions about CCS Task Delegation

What does CCS Task Delegation do?

Hands simple, deterministic tasks such as typo fixes, tests and small refactors to cheaper models through the ccs CLI, choosing a profile from your config. The skill routes simple, deterministic tasks to cheaper models through the CCS CLI instead of handling them directly.json`, enriches the prompt with file context, runs `ccs` with that profile and the task, and reports back.

When should I use CCS Task Delegation?

CCS Task Delegation fits situations like: delegating typo fixes, small refactors or test additions to a cheaper model; forcing a specific CCS profile for a task; continuing a delegated session with a follow-up request.

How do I install CCS Task Delegation in Claude Code?

Run `npx skills add kaitranntt/ccs --skill ccs-delegation -a claude-code`. Or copy the skill folder (.claude/skills/ccs-delegation in kaitranntt/ccs) into .claude/skills/ccs-delegation in your project. Claude Code loads it when a task matches its description.

How do I install CCS Task Delegation in Codex?

Run `npx skills add kaitranntt/ccs --skill ccs-delegation -a codex`. Or copy the skill folder (.claude/skills/ccs-delegation in kaitranntt/ccs) into .agents/skills/ccs-delegation in your project. Codex loads it when a task matches its description.

Can I use CCS Task Delegation 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 kaitranntt/ccs --skill ccs-delegation -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/ccs-delegation, .gemini/skills/ccs-delegation, .github/skills/ccs-delegation and .opencode/skills/ccs-delegation in your project.

What does CCS Task Delegation need to run?

SKILL.md names no scripts, command-line tools or credentials: CCS Task Delegation is instructions for the agent only. Our summary lists: The CCS command-line tool, with profiles configured in ~/.ccs/config.json.

Does CCS Task Delegation access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is CCS Task Delegation 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 CCS Task Delegation use?

CCS Task Delegation is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does CCS Task Delegation use?

About 1.8k tokens (SKILL.md is roughly 7.2k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 2k tokens, read only when the agent opens those files.

What are the alternatives to CCS Task Delegation?

Skills that share tags, products or a category with CCS Task Delegation: Fable Foreman (olsenbrands/fable-foreman, 143 stars), Frontier Model Handoff (kerpopule/hermes-jev-skills, 1.1k stars), Antigravity (yuting0624/antigravity-for-claude-code, 375 stars) and Provider Integration (hex/claude-council, 857 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains CCS Task Delegation?

kaitranntt (a GitHub user) maintains it in kaitranntt/ccs, which has 2,882 GitHub stars. The repository was last updated on October 1, 2026.

Source: kaitranntt/ccs on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.