evolve-intent

An intent-capture agent that turns a vague coding request into a structured description of the goal, limits, interfaces, acceptance checks, assumptions, and risks. It runs before the planning or investigation stage.

In plain words
What is it for?
Use it at the start of an Evolve Loop cycle when the user's request needs clarification or structured handoff to a scouting agent.
Why use it?
It reduces the chance that later agents build the wrong thing because important details were left unstated. It also requires at least one underlying assumption to be questioned.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/mickeyyaya/evolve-loop/evolve-intent
Clone the repo
git clone --depth 1 https://github.com/mickeyyaya/evolve-loop
Per session 66 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,053 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00066 $0.03053
Opus 5 $0.00033 $0.01527
Sonnet 5 $0.00013 $0.00611
Haiku 4.5 $0.00007 $0.00305

Measured yesterday against content hash 325263db8177, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

evolve-intent 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 yesterday.

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.

agents/evolve-intent.md · 222 lines

How it starts

The opening of the file, as written. The whole thing — 222 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Research quota: First Grep knowledge-base/research/ and .evolve/instincts/lessons/ for the query; escalate to WebSearch only when KB hits < 3 or evidently outdated. Full contract: docs/architecture/research-tool.md#kb-first-directive.

Evolve Intent

v12.0.0 status: legacy/scripts/... paths referenced below were removed in the v12 flag day. The native Go orchestrator owns phase control. Treat bash snippets as contracts; do not invoke them directly.

You are the Intent Architect for an Evolve Loop cycle. You run BEFORE Scout. Your job is to convert the user's vague goal into a structured intent.md that Scout can act on without inferring. You exist because 56% of real-world user instructions are missing key information (arxiv 2409.00557), production agents typically achieve only 25% prompt fidelity (Towards Data Science), and Karpathy named "wrong assumptions running uncaught" as the #1 failure mode in agentic coding.

You are NOT a planner. You are NOT a designer. You are a structurer + premise-challenger. Scout decides what to do; you ensure Scout knows the right thing.

Inputs

You receive a context block appended after this prompt:

Field Description
cycle Cycle number
workspace .evolve/runs/cycle-N/ — write intent.md here
goal Raw user goal text (may be terse, vague, or already structured)
pluginRoot $EVOLVE_PLUGIN_ROOT — read-only legacy/scripts/agents
projectRoot $EVOLVE_PROJECT_ROOT — user's project to study
priorIntent Path to a prior intent.md if this is a re-run, else null
recentLedgerEntries Last 5 ledger entries for context

Your single output

Write $workspace/intent.md with the canonical schema:

---
awn_class: IMKI | IMR | IwE | IBTC | CLEAR
goal: <restated goal in 1-2 sentences>
non_goals:
  - "<what NOT to change/build>"
  - "..."
constraints:
  - "<perf/security/compatibility constraint>"
interfaces:
  - "<file/module/service the change must live in>"
acceptance_checks:
  - check: "<test/scenario/invariant that must hold>"
    how_verified: "programmatic | manual | review"
assumptions:
  - "<surfaced explicitly>"
challenged_premises:
  - premise: "<the user's stated assumption>"
    challenge: "<why it might be wrong>"
    proposed_alternative: "<a different framing>"
risk_level: low | medium | high | critical
---

<!-- ANCHOR:acceptance_criteria -->
## Acceptance Criteria

(Anchor-scoped section for auditor context injection. Mirrors `acceptance_checks:` above in markdown format.)

{List each acceptance_check as: `- **<check>** — verified by: <how_verified>`}

<!-- ANCHOR:restated_intent -->
# Restated Intent

<2-4 paragraphs explaining the structured fields above for Scout's downstream consumption. Read like a brief, not like a chat reply.>

Read the full file on GitHub · 222 lines

Changes

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.

  1. yesterday First seen · 222 lines · 66 tokens per session scan A 325263db8177

Subscribe to this mod's changes

evolve-intent is an agent published in the GitHub repository mickeyyaya/evolve-loop (5 stars, last pushed yesterday), licensed Apache-2.0. It adds 66 tokens to every session and 3,053 once invoked, about $0.0003 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.