Writes the success criteria of an optimization or evaluation loop as one maximized metric, relaxable bounds on every other metric, and a written trade licence. Use when writing or revising goals.json, acceptance thresholds, benchmark targets, or agent objectives, especially with more than one metric, and when an iteration loop has stalled because every proposed change is rejected for regressing something. Not for running the iterations themselves (use eval-loop).
Installs into .claude/skills of the current project.
Are you the author of Bounded Goals?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/kohulan-bounded-goals)
---
name: bounded-goals
description: "Writes the success criteria of an optimization or evaluation loop as one maximized metric, relaxable bounds on every other metric, and a written trade licence. Use when writing or revising goals.json, acceptance thresholds, benchmark targets, or agent objectives, especially with more than one metric, and when an iteration loop has stalled because every proposed change is rejected for regressing something. Not for running the iterations themselves (use eval-loop)."
---
# bounded-goals
## The rule
**Maximize one metric. Give every other metric an explicitly relaxable bound, plus a
written trade licence.**
A conjunction of soft objectives — *maximize A and minimize B and minimize C* —
is an implicit veto on every Pareto move: it forbids progress.
## Why this exists
A naming-system project documented the failure directly: under the unbounded
multi-objective form, the agent reverted any change that regressed any
metric, and froze. Rewritten as one maximized metric with relaxable bounds, the
runtime ceiling was deliberately loosened 3.7 → 5.0 → 6.2 → 20.6 ms to let
accuracy move, then optimized back down to 10.2 ms at the end. Accuracy went
95.0% → 97.8% while mean name length *fell* 98.2 → 83.0 characters.
A sibling project's criteria read `G1 >= baseline AND G2 >= 0.45 AND G3 >= 0.95`, with no
trade licence: the freeze configuration. Bounded: maximize G1; G2 >= 0.45, relaxable to 0.40
if G1 gains a point; G3 >= 0.95, fixed.
## The contract (all five fields required; the layout is a default)
```
maximize: ONE metric — the thing the project is actually for
bounds: every other metric, each with a bound (floor or ceiling) marked relaxable or not
licence: plain sentences — "may regress X to Y in order to move Z"
held-out: a split the loop never reads; run by a human at intervals
stamp: derived:true on subset runs, so a subset number can never be
quoted as a headline result
```
## Rules of thumb
- **One maximand.** If two things both feel primary, the project has not decided
what it is for. Decide before optimizing.
- **Bounds are dials, not walls.** Write the relaxation policy at the same time
as the bound, or it will be treated as sacred at the worst moment.
- **Add a proxy bound for quality that the maximand cannot see.** Output length,
runtime per item, abstention rate. Cheap, and it catches degenerate wins.
- **Never rank on the proxy.** Bounding output length is healthy; *selecting*
candidates by length produces truncated garbage that still passes the oracle.
Bound it, do not optimize it.
- **Check the bounds can all hold before the loop starts.** The baseline must meet every
fixed bound, and no two bounds may forbid every move (coverage floor + byte-identical output).
- **Keep the held-out set genuinely held out.** Once it has informed a decision, it is a dev
set. Optional guard: the opt-in `guard-holdout` hook (`hooks/guard-holdout.py` in this
plugin; setup in `hooks/README.md`) stops a held-out run until the user has agreed.
- **Diagnostics ride free.** One expensive pass should emit status, outputs,
proxies, telemetry and failures together, so a new question costs nothing.
## When the loop stalls
A plateau is a signal about the *objective*, not the implementation. Before
trying harder:
1. Check whether a bound is doing the blocking, and whether it is relaxable.
2. Check whether the metric can still move at all — is there headroom, or is the
remaining failure mass a class the current architecture cannot express?
3. If neither, hand off: a reviewer defines a new objective, and a fresh
context implements it. Accumulated context is part of what plateaued.
## Red flags
| Thought | Reality |
|---|---|
| "All three metrics must not regress" | You have written a veto, not a goal. |
| "The bound is a hard requirement" | Then say who can relax it, and when. |
| "We improved on the subset" | Subset numbers are not headline numbers. Stamp them. |
| "Accuracy went up on the benchmark we tune against" | Of course it did. What did the held-out set say? |
| "Let's optimize the proxy too" | Proxies are bounds. Optimizing them is Goodhart's law with extra steps. |
