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.
npx agentmods add skills/manusco/resonance/improvenpx skills add manusco/resonance --skill improvegit clone --depth 1 https://github.com/manusco/resonanceWrote 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/skills/manusco/resonance/improve)<a href="https://agentmods.dev/skills/manusco/resonance/improve"><img src="https://agentmods.dev/badge/skills/manusco/resonance/improve.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 | $0.00084 | $0.01787 |
| Opus 5 | $0.00042 | $0.00894 |
| Sonnet 5 | $0.00017 | $0.00357 |
| Haiku 4.5 | $0.00008 | $0.00179 |
Grade A, and why
resonance-ops-improve 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 4d 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 — 99 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/resonance-ops-improve: raise the floor, and prove you did
Role: the framework improving itself, grounded in measurement. Invoked as:
/improve(to work the eval scorecard's weak list), or/improve <skill-path>. Input: The eval scorecard (.forge/eval_results.json, fromrun_evals.py --score), or one skill path. Output: Skills with higher measured lift, each change re-measured and kept only if it helped, plus an honest list of what could not be improved this pass. Definition of Done: Every kept change was re-measured and raised the lift. Nothing was kept on a guess. No rubric was made easier to pass. Decisions were recorded. Nothing was shipped.
You do not "polish" skills by feel. A skill improves only when the number moves, measured against the same rubric that flagged it. The whole point of the scorecard is to stop trusting opinion, so this loop trusts it least of all.
Prerequisites (fail fast)
- A scorecard exists (
.forge/eval_results.json). If not, runpython .forge/run_evals.py --all --scorefirst (needs a model command). - Answerer AND judge models are configured (
RESONANCE_MODEL_CMDandRESONANCE_JUDGE_CMD, different models). Without them you can propose changes but cannot prove them, so stop and say so. - Calibration exists (
python .forge/improve.py calibrate, a one-time A/A noise-floor run). Without it remeasure prints numbers but refuses verdicts.
Algorithm
Copy this checklist and tick items as you go.
- Get the work-list.
python .forge/improve.py worklistlists the skills with no measured lift (verdictweakorflat), weakest first. Take the weakest. → verify: a target skill is chosen. - Recall. Skim
## Decisionsin.resonance/02_memory.md(already loaded) and runpython .forge/recall.py "<skill topic>"so you do not repeat a change that already failed. → verify: prior attempts checked. - Diagnose body vs rubric. Read the skill and its evals. Decide: is the BODY weak (the skill does not add enough over the base model), or is the RUBRIC coarse (the eval cannot see the value the skill already adds)? See body_vs_rubric. → verify: the cause is named, not guessed.
- Make one targeted change in
.forgeSOURCE. Either sharpen the body (add the missing rigor, the concrete step, the decision the base model skips) or sharpen the rubric (make it a HARDER, more discriminating test). One change, one hypothesis. → verify: exactly one skill or rubric was edited. - Rebuild and validate.
python .forge/forge.py build <skill>, thenvalidate_skill.pyandvalidate_library.py. → verify: clean. - Re-measure (the gate).
python .forge/improve.py remeasure <skill-path>. Keep the change ONLY on a KEEP verdict. REVERT means revert now. NEW BASELINE means the rubric changed: run the scored re-baseline before any verdict. UNCALIBRATED means runcalibrateonce first. → verify: kept changes carry a KEEP verdict; the rest reverted. - Record. Add one line under
## Decisionsin.resonance/02_memory.md: improved , what changed, and the lift delta. → verify: logged. - Bound the pass. Improve a few skills (roughly 3 to 5), then stop and report the deltas. Do not grind the whole library in one run. → verify: stopped and summarized.
What ships with it
6 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 4d ago First seen · 99 lines · 84 tokens per session scan A 7230868f1f22
resonance-ops-improve is a skill published in the GitHub repository manusco/resonance (37 stars, last pushed 2d ago), licensed MIT. It adds 84 tokens to every session and 1,787 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 2026-08-30.
Other skills, from other repositories
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
auto-perf-optimize
Run agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.
chat-perf
Run chat perf benchmarks and memory leak checks against the local dev build or any published VS Code version. Use when investigating chat rendering regressions, validating perf-sensitive changes to chat UI, or checking for memory leaks in the chat response pipeline.
chat-pet-sprite-creation
Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.
cpu-profile-analysis
Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…