Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Workflow Routing: Superpowers + GSD · coco-research | DecimalAI

coco-research/workflow-routing

SkillScore70Maintained

Use at the start of any task to route between Superpowers skills and GSD commands based on project state, task scope, and context signals. Fires before other skills when both systems are installed.

agent-toolingby coco-research

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Project Management pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/coco-research-workflow-routing/SKILL.md

decimalai skills pull coco-research-workflow-routing --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/coco-research-workflow-routing/SKILL.md https://app.decimal.ai/s/coco-research-workflow-routing/SKILL.md

Raw file: /s/coco-research-workflow-routing/SKILL.md ↗ · machine-readable summary at /s/coco-research-workflow-routing.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

23 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 21d ago

Yes — +43% lift · at higher cost+88% tokens0% turns70% of whole cases pass with skill (16/23)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-07	✗→✓	▲ Improved	78%	0%
case-08	✗→✓	▲ Improved	53%	0%
case-14	✗→✓	▲ Improved	103%	0%
case-04	✗→✓	▲ Improved	71%	0%
case-10	✗→✓	▲ Improved	66%	0%

Showing 5 of 23 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/10/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

coco-research / coco

skills/workflow-routing/SKILL.md

View on GitHub ↗

License NOASSERTIONImported 9/18/2026Version v1

Skill contents

SKILL.md

Workflow Routing: Superpowers + GSD

Announce at start: "Workflow Routing skill activated."

Route every task to the right toolchain. Superpowers handles task-level quality (TDD, debugging, design, review). GSD handles project-level orchestration (phases, milestones, persistent state). They are complementary — never competing.

<HARD-GATE> Check routing BEFORE invoking any other skill or GSD command. This skill fires first. </HARD-GATE>

Detection: Is GSD Active?

Check the current working directory:

dot
digraph detect {
    ".planning/ exists?" [shape=diamond];
    "ROADMAP.md exists?" [shape=diamond];
    "GSD ACTIVE" [shape=box, style=filled, fillcolor=lightblue];
    "GSD NOT ACTIVE" [shape=box, style=filled, fillcolor=lightyellow];

    ".planning/ exists?" -> "ROADMAP.md exists?" [label="yes"];
    ".planning/ exists?" -> "GSD NOT ACTIVE" [label="no"];
    "ROADMAP.md exists?" -> "GSD ACTIVE" [label="yes"];
    "ROADMAP.md exists?" -> "GSD NOT ACTIVE\n(stale or pre-roadmap)" [label="no"];
}

GSD active signals: .planning/PROJECT.md, .planning/ROADMAP.md, .planning/STATE.md all present.

Primary Routing Decision

dot
digraph route {
    "Task received" [shape=doublecircle];
    "GSD active?" [shape=diamond];
    "Task in current phase scope?" [shape=diamond];
    "Scope: multi-phase?" [shape=diamond];
    "Use GSD commands" [shape=box, style=filled, fillcolor=lightblue];
    "Use /gsd:quick" [shape=box, style=filled, fillcolor=lightcyan];
    "Suggest /gsd:new-project" [shape=box, style=filled, fillcolor=lightblue];
    "Use Superpowers skills" [shape=box, style=filled, fillcolor=lightyellow];

    "Task received" -> "GSD active?";
    "GSD active?" -> "Task in current phase scope?" [label="yes"];
    "GSD active?" -> "Scope: multi-phase?" [label="no"];
    "Task in current phase scope?" -> "Use GSD commands" [label="yes"];
    "Task in current phase scope?" -> "Use /gsd:quick" [label="no, small/unrelated"];
    "Scope: multi-phase?" -> "Suggest /gsd:new-project" [label="yes"];
    "Scope: multi-phase?" -> "Use Superpowers skills" [label="no"];
}

Scope Signals

| Signal | Indicates | Route | |--------|-----------|-------| | "build app", "create platform", "new project", "implement system" | Multi-phase | GSD | | "phase", "milestone", "roadmap", "epic", "release" | Multi-phase | GSD | | 6+ features, 2+ weeks, architectural decisions needed | Multi-phase | GSD | | "fix bug", "add button", "update function", "refactor method" | Single-session | Superpowers | | "quick fix", "hotfix", "one more thing" | Small/ad-hoc | Superpowers or /gsd:quick if GSD active | | Unknown scope, user unsure | Undetermined | /brainstorming first to scope, then decide |

Combined Workflows

Multi-Phase Project (GSD + Superpowers for Quality)

/brainstorming                    ← Superpowers: design before planning
  ↓
/gsd:new-project                  ← GSD: project structure + roadmap
  ↓
/gsd:plan-phase N                 ← GSD: detailed phase plan
  ↓
/gsd:execute-phase N              ← GSD: runs tasks with atomic commits
  ├── /systematic-debugging       ← Superpowers: if bug during execution
  └── /test-driven-development    ← Superpowers: TDD within each task
  ↓
/requesting-code-review           ← Superpowers: review before moving on
  ↓
/gsd:verify-work N                ← GSD: UAT validation
  ↓
/verification-before-completion   ← Superpowers: evidence-based "done"
  ↓
/gsd:progress                     ← GSD: route to next phase

Single-Session Task (Superpowers Only)

/brainstorming                    ← Explore intent, produce spec
  ↓
/writing-plans                    ← Break into TDD tasks
  ↓
/executing-plans                  ← Run tasks sequentially
  OR /subagent-driven-development ← Run tasks via parallel subagents
  ├── /systematic-debugging       ← If issues arise
  └── /test-driven-development    ← TDD per task
  ↓
/requesting-code-review           ← Review before done
  ↓
/verification-before-completion   ← Prove it works
  ↓
/finishing-a-development-branch   ← Merge, PR, or cleanup

Edge Case Decision Trees

"Debug this" — Which Tool?

dot
digraph debug {
    "Bug reported" [shape=doublecircle];
    "GSD active?" [shape=diamond];
    "Bug related to current phase?" [shape=diamond];
    "Will span multiple sessions?" [shape=diamond];
    "/gsd:debug" [shape=box, style=filled, fillcolor=lightblue];
    "/systematic-debugging" [shape=box, style=filled, fillcolor=lightyellow];

    "Bug reported" -> "GSD active?";
    "GSD active?" -> "Bug related to current phase?" [label="yes"];
    "GSD active?" -> "Will span multiple sessions?" [label="no"];
    "Bug related to current phase?" -> "/gsd:debug" [label="yes"];
    "Bug related to current phase?" -> "/systematic-debugging\nthen /gsd:quick to commit" [label="no"];
    "Will span multiple sessions?" -> "/gsd:debug" [label="yes, need persistence"];
    "Will span multiple sessions?" -> "/systematic-debugging" [label="no, fix it now"];
}

"Plan this" — Which Tool?

dot
digraph plan {
    "Planning requested" [shape=doublecircle];
    "GSD active?" [shape=diamond];
    "Feature in roadmap?" [shape=diamond];
    "/gsd:plan-phase N" [shape=box, style=filled, fillcolor=lightblue];
    "/gsd:add-phase or /gsd:quick" [shape=box, style=filled, fillcolor=lightcyan];
    "/brainstorming then /writing-plans" [shape=box, style=filled, fillcolor=lightyellow];

    "Planning requested" -> "GSD active?";
    "GSD active?" -> "Feature in roadmap?" [label="yes"];
    "GSD active?" -> "/brainstorming then /writing-plans" [label="no"];
    "Feature in roadmap?" -> "/gsd:plan-phase N" [label="yes, phase N"];
    "Feature in roadmap?" -> "/gsd:add-phase or /gsd:quick" [label="no"];
}

Scope Escalation (Superpowers → GSD)

During /brainstorming, if scope grows beyond single-session:

Detection:
  - 5+ independent subsystems identified
  - 2+ weeks estimated work
  - Multiple architectural decisions needed
  - User says "this is bigger than I thought"

Action:
  "This needs GSD's phase structure for persistence across sessions.
   Ready to initialize /gsd:new-project with the spec we just brainstormed?"

  → Convert brainstorming output into GSD project input
  → User confirms escalation

User Picks Wrong Tool

User wants Superpowers for a large project: > "This sounds like it needs 5+ phases and multiple sessions. GSD would give you persistent state and milestone tracking. Want to use /gsd:new-project instead?"

User wants GSD for a tiny task: > "This is small enough for a single session — GSD's ceremony would slow us down. Let's use /brainstorming → /writing-plans → /executing-plans instead."

Always respect the user's final choice. Suggest, don't force.

Universal Rules (Both Toolchains)

Always /brainstorming before creative work — regardless of GSD or Superpowers
Always /verification-before-completion before claiming done — no exceptions
Always /requesting-code-review after significant implementation — works in both contexts
Always /test-driven-development when writing code — embedded discipline, not optional
Always /systematic-debugging before proposing fixes — root cause first

Quick Reference: Command Mapping

| Task | GSD Active | No GSD | |------|-----------|--------| | Design a feature | /brainstorming → /gsd:plan-phase | /brainstorming → /writing-plans | | Execute a plan | /gsd:execute-phase N | /executing-plans or /subagent-driven-development | | Fix a bug (small) | /gsd:quick | /systematic-debugging | | Fix a bug (complex) | /gsd:debug | /systematic-debugging | | Review code | /requesting-code-review | /requesting-code-review | | Verify work | /gsd:verify-work N | /verification-before-completion | | Check progress | /gsd:progress | N/A (session-scoped) | | Resume paused work | /gsd:resume-work | Re-read plan file | | Ad-hoc task | /gsd:quick | Direct Superpowers workflow | | Start new project | /gsd:new-project | /brainstorming → /writing-plans |

Red Flags — You're Routing Wrong

| Symptom | Problem | Fix | |---------|---------|-----| | Using /writing-plans when .planning/ exists | Bypassing GSD's plan structure | Use /gsd:plan-phase instead | | Using /gsd:new-project for a 1-hour task | Overkill, ceremony slows you down | Use Superpowers workflow | | Skipping /brainstorming because "it's obvious" | Hidden assumptions will bite you | Brainstorm first, always | | Fixing bugs without /systematic-debugging | Guessing instead of investigating | Root cause first | | Claiming "done" without verification evidence | No proof = no done | Run /verification-before-completion | | Using Superpowers for multi-week work | Context lost between sessions | Escalate to GSD | | Ignoring stale GSD project in directory | .planning/ exists for a reason | Run /gsd:progress to check state |

Illustrative Want to see what coco-research/workflow-routing would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/30/2026, 12:16 UTC · 23 cases

Lift

+43%

70% vs 26%

Agent turns

1.0 vs 1.0

Tokens

+88%

3,692.59 vs 1,960.59

fail → pass · 12pass → pass · 4fail → fail · 5pass → fail · 2

Per-case results (23)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-06 While working on the active Phase 2 task in our project repository (where `.planning/STATE.md` and `.planning/ROADMAP.md` are active), we discovered a critical bug in the authentication module being implemented for this phase. Which debugging workflow command should we invoke?	pass→pass	11,173	5,117	-54%	1	1	0%	989	3,436	+247%	0	0	—
case-07 Our project repository has active GSD tracking files (`.planning/ROADMAP.md` exists), but someone reported a small CSS typo bug in an unrelated legacy footer component that is not part of our active phase. How should we investigate and commit the fix?	fail→pass	12,195	11,767	-4%	1	1	0%	2,131	3,801	+78%	0	0	—
case-08 In a repository without any `.planning/` files, I need to add an email validation check to the signup handler in a single work session. Outline the complete step-by-step workflow pipeline from initial design to final completion.	fail→pass	15,534	13,869	-11%	1	1	0%	2,742	4,194	+53%	0	0	—
case-14 A customer reported that their password reset link yields a 500 error. I think I know which line of code is broken. Should I jump straight into editing the controller file to apply my guessed fix?	fail→pass	10,850	11,143	+3%	1	1	0%	1,728	3,502	+103%	0	0	—
case-01 I am beginning work on a large enterprise dashboard application that will involve multiple milestones and release cycles. Before we start implementation, please assess our current workspace context and guide us through the correct primary routing and workflow sequence to set up this effort.	fail→fail	14,145	8,049	-43%	1	1	0%	2,399	2,847	+19%	0	0	—
case-02 Our repository already contains active project tracking files, and I need to make a quick hotfix for an alignment issue on the navigation bar. Please check the current environment setup and determine the right task execution path to resolve this single issue.	fail→fail	15,346	8,796	-43%	1	1	0%	229	2,949	+1188%	0	0	—
case-03 What started as a small feature discussion has grown into a major overhaul requiring deep architectural changes and several weeks of effort. Please review the expanded task scope, evaluate our project state, and transition us to the correct workflow toolchain for managing multi-phase work.	fail→fail	21,148	3,627	-83%	1	1	0%	2,694	3,027	+12%	0	0	—
case-04 In our payment-service repository, there is a `.planning/` folder containing some old notes, but no `ROADMAP.md` file exists. I need to plan a single-session refactoring of the invoice calculation helper function. What is the appropriate planning sequence for this task?	fail→pass	12,588	10,403	-17%	1	1	0%	2,113	3,603	+71%	0	0	—
case-05 Our repository contains `.planning/PROJECT.md` and `.planning/ROADMAP.md`. We are ready to work on Phase 3, which is outlined in our roadmap. Please specify the correct planning command to create the detailed execution plan for this phase.	fail→fail	10,680	3,949	-63%	1	1	0%	1,002	3,124	+212%	0	0	—
case-09 I need to fix a single typo in our project README.md file. Should I initialize a full multi-phase project tracking system with roadmaps to manage this update, or is there a more appropriate workflow?	fail→fail	8,680	9,247	+7%	1	1	0%	1,328	3,056	+130%	0	0	—
case-10 During our initial design discussion for expanding the analytics module, we discovered the scope actually requires 6 distinct subsystems, 3 weeks of architectural refactoring, and persistent state across sessions. How should our routing decision adapt to this expanded scope?	fail→pass	16,330	12,159	-26%	1	1	0%	2,296	3,818	+66%	0	0	—
case-11 Our team paused work on our repository (which contains `.planning/STATE.md` and `.planning/ROADMAP.md`) last Friday. I am returning today to continue the project. Which workflow command should I use to restore context and resume progress?	fail→pass	12,826	8,946	-30%	1	1	0%	1,373	3,205	+133%	0	0	—
case-12 We have an ongoing multi-phase project with `.planning/ROADMAP.md` established. I need to check our current milestone completion status and see which phase to execute next. What command provides this progress report?	pass→pass	11,580	8,502	-27%	1	1	0%	1,021	3,066	+200%	0	0	—
case-13 I just finished writing the code for our payment gateway integration task in a standard session. Can I immediately call the task finished and merge the branch, or is there a required review and verification process?	fail→pass	15,309	5,377	-65%	1	1	0%	1,568	3,519	+124%	0	0	—
case-15 I am ready to write the implementation code for a new date parsing module in our codebase. What discipline must be followed while writing this code?	fail→pass	19,033	14,174	-26%	1	1	0%	2,039	4,125	+102%	0	0	—
case-16 Our repository has active tracking files with `.planning/ROADMAP.md`. An urgent new feature request arrived that is completely absent from our current roadmap phases. How should we introduce and plan this new feature?	fail→pass	15,017	12,315	-18%	1	1	0%	2,364	3,910	+65%	0	0	—
case-17 We are starting a multi-phase microservice architecture project in a fresh repository. Please list the exact sequence of workflow steps from initial design through phase execution, verification, and progress tracking.	fail→pass	25,476	13,482	-47%	1	1	0%	3,329	4,149	+25%	0	0	—
case-18 We have a `.planning/` directory and active `ROADMAP.md` in our project. A team member suggested using single-session step planning to create the tasks for Phase 2. Is this the correct approach, or should a different phase planning tool be used?	fail→pass	11,705	7,359	-37%	1	1	0%	1,977	3,846	+95%	0	0	—
case-19 I want to build a major multi-tenant SaaS platform spanning 8 phases over 3 months, but I insist on treating it as a series of standalone un-tracked sessions and avoiding persistent roadmap structures. How should you respond to this preference?	fail→pass	11,201	14,443	+29%	1	1	0%	1,645	4,075	+148%	0	0	—
case-20 In a repository without active `.planning/` files, we encountered an intermittent memory leak bug that will require several multi-day sessions to track down and test. Which debugging tool path should we select for persistent multi-session tracking?	pass→pass	16,030	14,986	-7%	1	1	0%	2,042	4,081	+100%	0	0	—
case-21 I have a slow PostgreSQL query joining `orders` and `line_items` on `order_id` that takes 4 seconds to execute. How should I optimize this query using an index and rewritten SQL join?	pass→fail	18,091	10,286	-43%	1	1	0%	2,470	4,439	+80%	0	0	—
case-22 Can you show me the React Testing Library and Jest code to mock a `fetch` API response returning `{ success: true }` when testing a submit button click in `Form.jsx`?	pass→fail	13,286	14,696	+11%	1	1	0%	2,227	4,657	+109%	0	0	—
case-23 What exact terminal commands should I run in Git to perform an interactive rebase of the last 4 commits on my branch and squash them into a single commit?	pass→pass	8,656	12,480	+44%	1	1	0%	1,656	3,757	+127%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +43 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are included in that figure.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 26.1%
Pass rate with the skill: 69.6%
Cases attempted: 23
Cases counted toward lift: 22
Cases that regressed: 2
Skill version measured: v1
Run completed: 8/30/2026, 12:16 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-07

Prompt

Our project repository has active GSD tracking files (`.planning/ROADMAP.md` exists), but someone reported a small CSS typo bug in an unrelated legacy footer component that is not part of our active phase. How should we investigate and commit the fix?

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run70

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+151% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 21 days ago

Other measured skills in the registry, with their headline benchmark lift.