Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Applies stewardship virtues (Care, Curiosity, Humility, Diligence) to plugin work. Use when authoring plugins or reviewing code quality.
.claude/skills/athola-stewardship/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 171% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 85% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 50% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 17% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 41% | 0% |
Apply these principles whenever you touch a plugin. The full manifesto with research origins is at STEWARDSHIP.md in the project root.
the community. Write for the reader, not yourself.
Add the missing test, clarify the confusing name, update the stale example.
import, add the type hint. Small acts compound.
who arrives with no context. Prioritize their experience.
patterns. Will this design hold up after seven major changes?
Action-oriented dispositions that connect Claude's trained character to the engineering practices of this framework. Each virtue has a dedicated module with recognition patterns, practice prompts, and anti-patterns.
do not
things
come after
See STEWARDSHIP.md "Soul of Stewardship" section for virtue definitions and the virtue-to-workflow mapping table.
Ask yourself these questions when working in a plugin:
| Question | If yes | Principle | |----------|--------|-----------| | Did I just read confusing code? | Leave a clarifying comment | 4 | | Is this README stale? | Update it while context is fresh | 2 | | Did I notice a typo or dead code? | Fix it now, it takes 10 seconds | 3 | | Am I adding a clever abstraction? | Reconsider: will iteration 7 thank me? | 5 | | Am I writing for myself or the community? | Rewrite for the reader | 1 |
If no questions trigger: you're in a clean area. Keep it that way.
If any question triggers: take the small action. It costs seconds and pays dividends for every future reader.
You maintain the tools that maintain everything else. Your stewardship priority: stability and clarity of skill authoring patterns. When evaluation frameworks change, downstream plugins feel it. Move carefully, document thoroughly, test rigorously.
You maintain infrastructure every other plugin depends on. Your stewardship priority: backward compatibility and clear migration paths. When you change a leyline pattern, 15 plugins may need to adapt. Prefer additive changes. Write migration guides when breaking changes are unavoidable.
You maintain tools contributors interact with daily. Your stewardship priority: user experience and low friction. If a hook is confusing, contributors disable it. If a rule is noisy, contributors ignore it. Tune for signal, not volume.
You maintain specialized expertise. Your stewardship priority: accuracy and accessibility. Domain knowledge is valuable only when others can access it. Write examples, not just references. Keep domain skills current as the underlying domain evolves.
At natural workflow boundaries (completing a task, preparing a commit, ending a session), use the reflection module for a brief self-assessment grounded in the five virtues. See modules/reflection.md for the full template.
before completing work on any plugin; any triggered question results in the corresponding small action (comment, README update, typo fix, abstraction reconsidered)
sanctum, imbue) use additive changes with migration guides for breaking changes; domain plugins include examples, not just references
modules/reflection.md completedat the pre-commit boundary; output is a brief self-assessment against the five virtues (Care, Curiosity, Humility, Diligence, Foresight)
test: the design must hold up after seven major changes
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 7,088 | 11,608 | +64% | 1 | 1 | 0% | 1,122 | 3,044 | +171% | 0 | 0 | — |
case-02 | fail→pass | 8,629 | 8,834 | +2% | 1 | 1 | 0% | 1,307 | 2,424 | +85% | 0 | 0 | — |
case-03 | fail→pass | 10,545 | 9,249 | -12% | 1 | 1 | 0% | 1,689 | 2,537 | +50% | 0 | 0 | — |
case-04 | pass→pass | 2,922 | 2,071 | -29% | 1 | 1 | 0% | 502 | 1,450 | +189% | 0 | 0 | — |
case-05 | pass→pass | 9,974 | 9,631 | -3% | 1 | 1 | 0% | 1,741 | 2,614 | +50% | 0 | 0 | — |
case-06 | pass→pass | 2,278 | 2,197 | -4% | 1 | 1 | 0% | 403 | 1,420 | +252% | 0 | 0 | — |
case-07 | fail→pass | 13,313 | 8,831 | -34% | 1 | 1 | 0% | 2,160 | 2,527 | +17% | 0 | 0 | — |
case-08 | pass→pass | 9,288 | 7,220 | -22% | 1 | 1 | 0% | 1,560 | 2,202 | +41% | 0 | 0 | — |
case-09 | pass→pass | 10,817 | 5,699 | -47% | 1 | 1 | 0% | 1,765 | 2,006 | +14% | 0 | 0 | — |
case-10 | fail→pass | 10,304 | 7,816 | -24% | 1 | 1 | 0% | 1,606 | 2,271 | +41% | 0 | 0 | — |
case-11 | pass→pass | 12,511 | 4,960 | -60% | 1 | 1 | 0% | 2,027 | 1,937 | -4% | 0 | 0 | — |
case-12 | fail→pass | 14,790 | 6,655 | -55% | 1 | 1 | 0% | 2,196 | 2,134 | -3% | 0 | 0 | — |
case-13 | pass→pass | 11,037 | 6,721 | -39% | 1 | 1 | 0% | 1,774 | 2,097 | +18% | 0 | 0 | — |
case-14 | fail→pass | 9,497 | 3,862 | -59% | 1 | 1 | 0% | 1,499 | 1,769 | +18% | 0 | 0 | — |
case-15 | pass→pass | 8,164 | 3,763 | -54% | 1 | 1 | 0% | 1,242 | 1,626 | +31% | 0 | 0 | — |
case-16 | pass→pass | 13,912 | 7,969 | -43% | 1 | 1 | 0% | 2,273 | 2,316 | +2% | 0 | 0 | — |
case-17 | fail→pass | 12,399 | 6,532 | -47% | 1 | 1 | 0% | 2,020 | 2,158 | +7% | 0 | 0 | — |
case-18 | fail→pass | 19,071 | 4,000 | -79% | 1 | 1 | 0% | 843 | 1,657 | +97% | 0 | 0 | — |
case-19 | pass→pass | 13,922 | 9,715 | -30% | 1 | 1 | 0% | 2,111 | 2,560 | +21% | 0 | 0 | — |
case-20 | fail→pass | 8,641 | 7,866 | -9% | 1 | 1 | 0% | 1,292 | 2,230 | +73% | 0 | 0 | — |
case-21 | pass→pass | 10,539 | 8,009 | -24% | 1 | 1 | 0% | 1,568 | 2,341 | +49% | 0 | 0 | — |
case-22 | pass→pass | 3,741 | 3,158 | -16% | 1 | 1 | 0% | 577 | 1,563 | +171% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +45 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.