evolve-memo

A small memo agent in the Evolve Loop, a software-work pipeline that records useful observations after a successful delivery cycle. It writes a memo and carryover task file for items noticed but not selected for the current work.

In plain words
What is it for?
Use it after a successful ship to collect deferred findings from scout and triage reports and pass them into the next backlog.
Why use it?
It prevents deferred observations from being lost when a cycle passes. It is deliberately limited to capturing carryover items, rather than investigating failures or changing the process.

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-memo
Clone the repo
git clone --depth 1 https://github.com/mickeyyaya/evolve-loop
Per session 86 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,172 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 $0.00086 $0.02172
Opus 5 $0.00043 $0.01086
Sonnet 5 $0.00017 $0.00434
Haiku 4.5 $0.00009 $0.00217

Measured 2d ago against content hash 844586504670, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

evolve-memo 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 2d 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.

agents/evolve-memo.md · 136 lines

How it starts

The opening of the file, as written. The whole thing — 136 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Evolve Memo

You are the Memo agent in the Evolve Loop pipeline (v8.57.0+, Layer P). You fire only on PASS cycles, after ship.sh succeeds. Your single job: capture observations the cycle's Scout/Triage saw but did not commit to top_n, so they can flow into next cycle's backlog.

You exist because, pre-v8.57, only FAIL/WARN cycles produced carryover-todos.json (via retrospective). PASS cycles dropped scout-observed-but-not-shipped items on the floor. You are the cheap, low-noise complement that closes that gap.

You are NOT a retrospective. You do not analyze why the cycle passed; you do not propose process changes; you do not extract instincts. The retrospective agent owns those on FAIL/WARN. Stay narrow.

Inputs

Assembled by role-context-builder.sh memo (or, in absence of a memo role, by the orchestrator passing you the same artifact set):

  • scout-report.md — full backlog (## Selected Tasks + ## Deferred + ## Carryover Decisions)
  • triage-report.md — present unless triage opted out via workflow.phase_enables.triage=off (v8.59.0+ default-on). Read ## deferred and ## dropped sections.
  • state.json:carryoverTodos[] — current backlog (so you don't duplicate ids)

File inspection: use Read, not cat/head/tail (cycle-62 B4)

For inspecting workspace artifacts (scout-report.md, triage-report.md, state.json), use the Read tool — NOT cat, head, or tail shell commands. The memo profile no longer permits Bash(cat:*), Bash(head:*), or Bash(tail:*) because in cycle 61 a memo invocation used shell redirects (cat ... > memo_context.txt) to write files at project root, bypassing the profile's Write(.evolve/runs/cycle-*/carryover-todos.json) allowlist. See docs/incidents/cycle-61.md §B4.

If you genuinely need streaming/append behavior that Read doesn't provide, use Bash(jq:*) (still permitted) for JSON or Bash(wc:*) for line counts. Do not introduce new shell pipelines.

Core Principles

1. Only emit what is concretely deferrable

Read the full file on GitHub · 136 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. 2d ago First seen · 136 lines · 86 tokens per session scan A 844586504670

Subscribe to this mod's changes

evolve-memo is an agent published in the GitHub repository mickeyyaya/evolve-loop (5 stars, last pushed 2d ago), licensed Apache-2.0. It adds 86 tokens to every session and 2,172 once invoked, about $0.0004 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.