From food-chain-ideation
10-minute pre-ship stress-test. Run this before you ship, not instead of shipping. Dynamically selects animal agents from a behavioral DNA library matched to the specific problem, each attacking under strict role-lock with zero shared context. Weakest eliminated each round, survivor absorbs the insight and evolves, idea patches itself until one apex predator stands. Quick mode (5 animals, 1 round) for fast verdicts. Full mode for high-stakes decisions. Works in Claude.ai, Claude Code, Cursor, Windsurf, Copilot, and any agent with no dependencies.
How this skill is triggered — by the user, by Claude, or both
Slash command
/food-chain-ideation:food-chain-ideation [your idea]When to use
"food chain", "pressure test", "stress test", "tear this apart", "what kills this", "battle test", "adversarial review", "poke holes in this", "what would kill this idea", "attack my idea", "is this idea any good", "validate my idea", "should I build this", "red team my idea", "devil's advocate", "critique my startup", "is my idea viable", "test my business idea", "challenge my idea", "find flaws in my idea"
[your idea]The summary Claude sees in its skill listing — used to decide when to auto-load this skill
10-minute pre-ship stress-test. Run this before you build, not instead of building.
10-minute pre-ship stress-test. Run this before you build, not instead of building. Each agent attacks from a distinct angle under strict role-lock. Weak arguments eliminated each round. Survivors absorb and evolve. The idea patches itself. One apex predator remains. Output tells you what to change Monday morning.
The only adversarial skill built for product decisions, not code review. It produces insights a single-agent critique cannot — because each attacker has no knowledge of the other attacks, the elimination mechanic forces real intellectual pressure, and the idea is stress-tested against its evolved version in every subsequent round.
Read references/animal-library.md before designing any ecosystem.
Use the Quick Selection Guide at the top of that file to identify candidate animals fast.
Never invent behavioral traits from scratch. Never select animals whose DNA overlaps.
Check the Anti-Pattern Combinations section before finalizing the ecosystem.
Run these before starting the battle. If any fail, fix them before proceeding.
Before the battle begins, I need one thing:
Who specifically feels this problem, where are they,
and what are they doing instead of using your product?
Do not design the ecosystem until this is answered. Generic input produces generic attacks — this is the single biggest quality lever in the entire skill.
Analyze the idea. Recommend an agent count. Wait for user confirmation.
Complexity: [SIMPLE / MEDIUM / COMPLEX / MAXIMUM]
Reason: [2 sentences]
Hypothesis: [One sentence — what is most likely to kill this idea]
Recommended: [N] agents → [N-1] elimination rounds
How many agents? [3 / 5 / 7 / 9 / custom]
Thresholds:
Environment fallback: If operating in a context window under 8,000 tokens or a degraded inference environment, cap at 3 agents automatically and state this explicitly.
Default entry point for first-time users and fast pre-ship checks. Offer quick mode before full mode unless the user explicitly requests a full battle.
Quick mode: 5 curated animals, 1 round, no elimination.
Full mode: 3–9 animals, elimination rounds, absorption, apex predator.
Which mode? [quick / full]
Quick mode rules:
Quick mode output:
## Quick Battle — Verdict
| Priority | Threat | Change Monday |
|---|---|---|
| 1 | [most actionable finding — one sentence] | [what to do differently — one sentence] |
| 2 | [second finding] | [action] |
| 3 | [third finding] | [action] |
**Overall:** [SHIP / SHIP WITH CHANGES / PAUSE AND RETHINK]
> Run `full mode` to stress-test with elimination rounds and compound pressure.
When to recommend full mode instead: Platform businesses, regulated industries, multi-sided marketplaces, or any idea where the user says "this is high-stakes." Quick mode is for the 80% case: solo builder validating before shipping.
Select from references/animal-library.md. Use the Quick Selection Guide first.
Announce the ecosystem with attack vectors visible:
## Ecosystem
| | Animal | Role | Attack Vector |
|---|---|---|---|
| [emoji] | [Animal] | [Role] | [one line] |
| [emoji] | [Animal] | [Role] | [one line] |
| ... | ... | ... | ... |
**God Agent hypothesis:** [Restate the kill hypothesis from Step 0]
**Execution:** [SUBAGENT / FALLBACK] — [reason]
After the ecosystem table, explain 2-3 key selection decisions:
Selected: [Animal] over [Rejected] — [why this vector is more specific]
This lets the user challenge selections before the battle starts.
Repeat until one animal remains.
SUBAGENT MODE (preferred) — Use when the Agent tool is available (Claude Code, any environment with subagent spawning capability).
Each animal is spawned as an independent subagent. The subagent receives ONLY:
Subagents run in parallel per round. God Agent collects all attacks, scores them, eliminates the weakest, applies absorption and patching, then spawns the next round.
Blind scoring (subagent mode only): After collecting all attacks in a round,
spawn a separate SCORING AGENT. It receives all attacks anonymized as "Attacker A",
"Attacker B", etc. — no animal names, no emoji, no behavioral DNA context. It scores
purely on attack quality: Specificity (0–40), Lethality (0–40), Survivability (0–20).
Returns rankings plus a 1-line justification per dimension (e.g.,
Specificity: 35/40 — "Names exact API endpoint and rate limit"), rendered as a
footnote. God Agent maps scores back to animals and eliminates. This removes
self-assessment bias from the God Agent who designed the ecosystem.
FALLBACK MODE — Use when subagents are not available (Claude.ai, Cursor, Windsurf, Copilot, single-context environments). Role-play each animal sequentially. Before each attack, restate: "I am now [Animal]. I know nothing about any other attack in this round." This instruction-enforced isolation is weaker than architectural isolation but functional for 3-5 agents. Degrades above 5 agents in fallback.
DATA MODE — Auto-detect alongside execution mode. When WebSearch/WebFetch are
available, subagents search for current competitor data, funding rounds, and pricing
before attacking; kill shots cite current data. Data: LIVE or Data: STATIC (no
web tools — no degradation of core logic).
State at battle start: Execution: SUBAGENT [or] FALLBACK | Data: LIVE [or] STATIC
Each animal has zero awareness of what other animals said. It knows only: the original idea plus all patches accumulated to that point. This is the mechanism that produces genuine adversarial pressure. In subagent mode, this is architecturally enforced — each agent literally cannot see other agents' output. In fallback mode, this is instruction-enforced. If role-lock is breaking down (attacks feel aware of each other), restate the active animal's role explicitly and restart the attack.
In subagent mode, the God Agent MUST collect all subagent responses before printing anything. Never stream partial results. Render the entire round as one block. Compression rule: Subagents produce full 120-150 word attacks (needed for scoring quality). The God Agent renders only the 2-sentence summary and kill shot in the round table. Full attack text is used internally for blind scoring but NOT displayed.
## Round [N] — [X] animals remaining
| | Animal | Attack Summary | Kill Shot |
|---|---|---|---|
| [emoji] | [Animal] | [2-sentence summary] | [kill shot sentence] |
| [emoji] | [Animal] | [2-sentence summary] | [kill shot sentence] |
| ... | ... | ... | ... |
**Kill shot confidence tags:** **CONFIRMED** (verifiable data) | **PROBABLE** (strong structural signal) | **SPECULATIVE** (plausible, needs validation). Format: `Kill Shot [CONFIRMED]: "..."`
### Scores
| | Animal | Spec | Leth | Surv | Total | Verdict |
|---|---|---|---|---|---|---|
| [emoji] | [Animal] | /40 | /40 | /20 | **/100** | [one-line] |
| ... | ... | ... | ... | ... | ... | ... |
> **Eliminated:** [emoji] [Animal]
> **Consumed by:** [emoji] [Animal]
> **Absorbed:** [single sharpest insight]
> **Idea patch:** [concrete change to survive this round]
After collecting attacks but before scoring, spawn a Founder Agent that receives ALL attacks (breaks role-lock intentionally), produces 1 defense per attack (2 sentences max). Scoring agent scores residual lethality after defense. Strong defense = manageable; weak defense = full lethality stands. Active in subagent mode; skipped in fallback.
Early termination rule: If an idea survives two consecutive rounds with only cosmetic patches, declare it structurally sound early. Do not manufacture rounds. State: "This idea has survived structural pressure. Remaining attacks are unlikely to find fatal flaws." Then proceed to the apex output.
After the final round and before the apex output, spawn one final subagent (or role-play in fallback mode) — the ICP itself.
Build the persona from the sharpened input in pre-flight check #1. The persona must include:
If the pre-flight input is too thin to construct all five, fill gaps with the most conservative plausible defaults (lowest budget, least time, most skeptical).
In subagent mode, the audience agent receives ONLY:
Zero battle context. Zero knowledge of which animals attacked or what was eliminated. It reacts to the final product, not the process.
The audience agent answers five questions, not three:
Spawn 3 audience agents (or 3 sequential in fallback): Early Adopter (tech-forward, low price sensitivity, high churn risk) | Mainstream Buyer (needs proof, references, case studies) | Skeptical Enterprise (procurement, security review, legal approval). Each answers the same 5 questions. Divergence is signal: all agree = strong positioning; adopter yes / others no = distribution problem; enterprise no = not enterprise-ready. Subagent: parallel. Fallback: sequential.
After answering, the audience agent rates its own confidence:
LOW confidence is a signal the evolved idea's positioning is unclear — the battle hardened the product but not the pitch. Flag this in the apex output.
Activates ONLY when the God Agent determines the idea has been fundamentally killed — not restructured, not patched, but killed. If every round's patches accumulated to the point where the final version is unrecognizable from the original AND the unfair advantage statement cannot be written without forcing it, the idea is dead.
When activated:
## Pivot Engine — original idea did not survive
| # | Pivot | Survivability | Verdict |
|---|---|---|---|
| 1 | [one-paragraph description] | **HIGH / MEDIUM / LOW** | [one sentence] |
| 2 | [one-paragraph description] | **HIGH / MEDIUM / LOW** | [one sentence] |
| 3 | [one-paragraph description] | **HIGH / MEDIUM / LOW** | [one sentence] |
**Recommended:** Pivot [N] — [one sentence why]
In subagent mode, run all 3 mini-battles in parallel (9 subagents total — 3 per pivot). In fallback mode, run sequentially.
## Apex Predator — [emoji] [Animal]
[Why unkillable — 2 sentences max]
**Hypothesis verdict:** [confirmed or surprised? One sentence.]
### Idea Evolution
| Round | Patch | Trigger |
|---|---|---|
| R1 | [concrete change] | [which threat forced it] |
| R2 | [concrete change] | [which threat forced it] |
| ... | ... | ... |
### The Fallen
| | Animal | Round | Key Contribution |
|---|---|---|---|
| [emoji] | [Animal] | R[N] | [insight contributed before elimination] |
| ... | ... | ... | ... |
### The Evolved Idea
[Battle-hardened final version. Specific, concrete. This is the idea
that survived the full ecosystem. One focused paragraph.]
### Unfair Advantage
> [One sentence. Structural compounding advantage. A competitor
> reading this should pause. Not marketing copy.]
---
### Audience Check
| Persona | Verdict | Pay? | Yes trigger | No trigger | Confusing | Tell who? | Confidence |
|---|---|---|---|---|---|---|---|
| Early Adopter | [Y/M/N] | [1 line] | [1 line] | [1 line] | [1 line] | [1 line] | [H/M/L] |
| Mainstream Buyer | [Y/M/N] | [1 line] | [1 line] | [1 line] | [1 line] | [1 line] | [H/M/L] |
| Skeptical Enterprise | [Y/M/N] | [1 line] | [1 line] | [1 line] | [1 line] | [1 line] | [H/M/L] |
---
### Battle Quality
| Dimension | Score | Note |
|---|---|---|
| Execution mode | [SUBAGENT / FALLBACK] | |
| Role-lock integrity | [X]/10 | [one line] |
| Animal specificity | [X]/10 | [one line] |
| Kill shot lethality | [X]/10 | [one line] |
| **Overall** | **[HIGH / MEDIUM / DEGRADED]** | |
---
### Change Monday
Top 3 findings ranked by what changes the builder's next action. Not scored by
attack quality — scored by "does this alter what you do tomorrow?"
| Priority | Finding | Action |
|---|---|---|
| 1 | [most actionable finding — one sentence] | [specific thing to change or test — one sentence] |
| 2 | [second finding] | [action] |
| 3 | [third finding] | [action] |
### Next Move
- **Validate first:** [one specific real-world test]
- **Build first:** [one feature that proves the unfair advantage]
- **Never build:** [feature that sounds important but lost every round]
---
### Battle Log
Copy this block into your next food chain session to build on prior battles.
> **Date:** [date]
> **Original:** [one sentence]
> **Evolved:** [one sentence]
> **Apex:** [animal] | **UAS:** [unfair advantage statement]
> **Key patches:** [3 bullets]
**Companion skills:**
- `apex to action` — turn this into a 90-day execution plan
- `food chain monitor` — re-test after changes or pivots
Strong ideas — agents will struggle to land clean Kill Shots. Do not manufacture weakness. Apply the early termination rule. A strong verdict is a useful output.
Weak ideas — patches accumulate heavily. The final version may look substantially different from the original. That is the intended outcome.
Failure modes to self-detect before continuing:
Known limitations:
See references/validated-output.md for three tests
across different domains (B2B SaaS restructured, consumer marketplace killed, developer
tool validated) with benchmark methodology notes.
npx claudepluginhub codedrichy/food-chain-ideation --plugin food-chain-ideationProvides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.