outcome

A command for marking whether stored advice worked, failed, or helped only partly. It records that result against a memory item.

In plain words
What is it for?
For reviewing memory items after using them and recording outcomes such as worked, failed, or partial.
Why use it?
It helps separate reliable advice from advice that was wrong or incomplete, instead of treating every stored memory as equally trustworthy.

Command for Claude Code

Part of the memory-layer plugin — 3 skills, 6 commands, 3 hooks, 1 MCP server shipped together

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 commands/runtimenoteslabs/memory-layer/outcome
Clone the repo
git clone --depth 1 https://github.com/runtimenoteslabs/memory-layer

Made for: Claude Code.

Or install memory-layer, the plugin that ships this one along with the rest of its 3 skills, 6 commands, 3 hooks, 1 MCP server.

Per session 14 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 498 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 $0.00014 $0.00498
Opus 5 $0.00007 $0.00249
Sonnet 5 $0.00003 $0.00100
Haiku 4.5 $0.00001 $0.00050

Measured 3d ago against content hash 2de2b3eb7c4a, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

outcome 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 3d 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.

.claude/commands/outcome.md · 73 lines

What it actually says

Outcome Command

Record feedback on whether a memory's advice was helpful. This is the core of outcome-based learning - memories that help get boosted, memories that fail get penalized.

Usage

/outcome <memory_id> worked|failed|partial

Outcome Types

Outcome Score Change Use When
worked +0.2 The advice solved the problem
failed -0.3 The advice was wrong or unhelpful
partial +0.05 The advice was on the right track but incomplete

The asymmetric scoring (-0.3 for failed vs +0.2 for worked) is intentional: wrong advice wastes more time than good advice saves, so it should sink faster.

Examples

# Memory advice solved the problem
/outcome mem_abc123 worked

# Memory advice was wrong
/outcome mem_def456 failed

# Memory advice partially helped
/outcome mem_ghi789 partial

Implementation

Record the outcome:

mem outcome "$MEMORY_ID" "$OUTCOME"

Score Ranges

Memories have an outcome score from -1.0 to 1.0:

Score Range Meaning Retrieval Impact
> 0.5 Highly reliable Prioritized in results
0.3 to 0.5 Generally helpful Normal ranking
0.0 to 0.3 Mixed results Normal ranking
-0.3 to 0.0 Questionable Deprioritized
< -0.3 Unreliable May be auto-archived

Impact Over Time

With consistent feedback:

  • Week 1: ~70% retrieval precision (baseline)
  • Week 4: ~78% precision (some outcome data)
  • Week 8: ~85% precision (outcome scoring active)
  • Week 12: ~90% precision (bad memories suppressed)

Tips

  • Provide feedback soon after using a memory
  • Be honest about partial successes
  • The outcome-feedback Skill will prompt you naturally after solutions
  • Feedback on frequently-used memories has the most impact
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. 3d ago First seen · 73 lines · 14 tokens per session scan A 2de2b3eb7c4a

Subscribe to this mod's changes

outcome is a command published in the GitHub repository runtimenoteslabs/memory-layer (10 stars, last pushed 3mo ago), licensed MIT. It adds 14 tokens to every session and 498 once invoked, about $0.0001 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.