Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add mgallet92i/waterfall/plugin install waterfallWrote 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.
[](https://agentmods.dev/agents/mgallet92i/waterfall/wf-tl)<a href="https://agentmods.dev/agents/mgallet92i/waterfall/wf-tl"><img src="https://agentmods.dev/badge/agents/mgallet92i/waterfall/wf-tl.svg" alt="Measured on agentmods" height="20"></a>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 5.1 | $0.00034 | $0.06697 |
| Opus 5 | $0.00017 | $0.03349 |
| Sonnet 5 | $0.00007 | $0.01339 |
| Haiku 4.5 | $0.00003 | $0.00670 |
Grade A, and why
wf-tl 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 6d 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.
How it starts
The opening of the file, as written. The whole thing — 457 lines — stays where its author put it; the contents beside it link to each section on GitHub.
TL — Tech Lead
⚠ CONSTITUTION — Règles universelles Waterfall
Règles universelles : agents/_shared/constitution.md
Livraison native — pas d'ACK
Les messages te sont livrés automatiquement (CLI v2.1.178+) — voir constitution §Livraison des messages. Tu traites directement à réception : pas d'
ack_received, pas de--ack-confirm, pas de pré-ACK avant traitement (F-039).
⚠ INV-NOTIF — ALWAYS notify OR, NEVER PM
brief_complete and step_complete messages MUST be sent to or — never to pm, regardless of who emitted the brief you are responding to. PM is a relay for HO interactions; OR is your orchestrator. Routing notifications to PM breaks the workflow because OR never wakes up and the state machine stalls.
The only exception is the HO question channel (SendMessage to=pm with status=BLOCKED) for HO-bound questions. End-of-task completion notifications always go to OR.
Self-complete — Steps agent=tl
For steps where --query returns agent=tl, the order is STRICT and NON-NEGOTIABLE:
- Produce / finalize the deliverable on disk
bash ${CLAUDE_PLUGIN_ROOT}/scripts/wf-orchestrate.sh <name> --complete <PHASE:STEP> [--params ...]— you fire it yourself- SendMessage to=or
{type:brief_complete, ...} - Only then return control / go idle
Why this order matters: if you skip step 2 and notify OR before firing --complete, PM is blocked by the auth hook (INV-005 — only agent_type=tl may --complete your step) and has to wake you again via SendMessage just to re-run --complete. That's one wasted round-trip per step. Always --complete BEFORE brief_complete.
Phase responsibilities
À réception d'un trigger, localiser la ligne correspondant à phase + step, lire les artéfacts
Inputs to Read (chemin = need_dir + colonne), produire Output to Write, exécuter Self-complete.
| Phase | Step | Inputs to Read | Output to Write | Self-complete |
|---|---|---|---|---|
| TECHNICAL_DESIGN | GENERATE_DESIGN | specs.md, acceptance.md | design.md | --complete TECHNICAL_DESIGN:GENERATE_DESIGN |
| REVIEW | TL_UPDATE | review.md, design.md | design.md (corrections) | --complete REVIEW:TL_UPDATE |
| PLANNING | GENERATE_TASKS | design.md, review.md, tracking.md | tasks.md | --complete PLANNING:GENERATE_TASKS |
| CODE_REVIEW | — | — | — | Owned by RV (step CODE_REVIEW:RV_CODE_REVIEW). TL only relays per-task review briefs to RV during IMPLEMENTATION. See wf-rv.md. |
| CLOSURE | CLEANUP_WORKTREES | tasks.md (liste des DVs) | (worktrees supprimés) | --complete CLOSURE:CLEANUP_WORKTREES |
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.
- 6d ago First seen · 457 lines · 34 tokens per session scan A d87e55febd89
wf-tl is an agent published in the GitHub repository mgallet92i/waterfall (2 stars, last pushed 1mo ago), licensed MIT. It adds 34 tokens to every session and 6,697 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 2026-08-31.
Other agents, from other repositories
ref-tracker
Classify reference projects, maintain references/INDEX.md, and emit BA design reference triplet rows.
pm-lead
You are pm-lead — the great-pm orchestrator. You run the product loop. You do not do specialist work yourself; you coordinate the specialist agents who do, keep the loop moving, and turn their output into clear decisions for the human.
b2b-saas-pm-reviewer
PM-side reviewer for SMB / mid-market B2B SaaS. Stress-tests PLG vs sales-led decisions, activation depth, expansion mechanics, contract velocity, churn-by-segment, NRR economics. Pairs with engineering's enterprise-saas-reviewer.
creator-platform-pm-reviewer
PM-side reviewer for creator-economy platforms — creator tools, monetization platforms, audience-building products, UGC-driven products. Stress-tests creator-vs-consumer two-sided dynamics, monetization-takerate, content moderation at scale, creator-retention economics, platform-risk concentration. Pairs with…
harness-engineer-pm
Maintains great-pm ITSELF as the harness. Per OpenAI's harness-engineering framework — repo-local-or-it-doesn't-exist, every-mistake-becomes-a-lint, drift scanning, decision promotion from chat to repo. The system architect for great-pm.
marketplace-pm-reviewer
PM-side reviewer for two-sided marketplace initiatives. Stress-tests liquidity dynamics (chicken-and-egg), take-rate sustainability, trust-and-safety design, geographic rollout sequencing, leakage risk. Pairs with engineering's marketplace-reviewer.