Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Boundary Analysis Campaign — probe where methods fail, map validity envelopes, test robustness, catalog failure modes, detect scaling limits. 5 strategies, 3 tactics, 11 subagent SOPs.
.claude/skills/yogsoth-ai-boundary-analysis/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 388% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 42% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -1% | 0% |
| case-06 | ✓→✗ | ▼ Worse | -53% | 0% |
Probe where methods fail — validity envelopes, robustness, failure modes, scaling limits.
This campaign is a strategy book — CC reads, internalizes, and autonomously constructs an approach. The SKILL.md files are textbooks, not scripts.
| Signal | Strategy | |--------|----------| | Valid conditions, applicability range, multi-axis variation, degradation curves | → validity-envelope-mapping | | Cross-method consistency, multi-model convergence, Wimsatt robustness | → robustness-testing | | System boundaries, inclusion/exclusion, CSH critique | → boundary-critique | | Failure modes, boundary inputs, distribution shift, adversarial | → failure-mode-analysis | | Scaling limits, scaling law, behavioral phase transitions, capacity | → scaling-frontier |
Import (5): web-search, web-research, paper-overview, paper-search, paper-research
Subagent (11): assumption-enumeration, alternative-model-generation, convergence-assessment, fragility-flagging, variation-axis-definition, controlled-perturbation, validity-envelope-construction, edge-case-generation, failure-clustering, scaling-regime-detection, boundary-synthesis
Shared (1): evidence-synthesis
context-checkpoint after each strategy completes.
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Strategy | When to use | | --- | --- | | boundary-critique | Apply CSH boundary critique — what is included/excluded, who benefits/is harmed, what expertise is privileged/marginalized. Identifies opportunities at the boundaries. | | deep-insight-validity-envelope-mapping | Map multi-dimensional validity envelopes — define variation axes, perturb systematically, measure degradation, construct boundary surface. | | failure-mode-analysis | Systematically catalog failure modes — generate edge cases, observe failures, cluster by mechanism, identify triggers and frequency. | | robustness-testing | Test conclusion robustness via multi-model convergence — enumerate assumptions, generate alternatives, compare results, flag fragile conclusions. | | scaling-frontier | Analyze behavior across scales — detect regime changes, identify capacity limits, fit scaling laws within regimes. |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | boundary-synthesis | Compile all boundary analysis products into a coherent report — validity envelopes, robustness results, failure catalogs, scaling maps, safe operating conditions. | | context-checkpoint | Append research process and results to the current Phase's context file. Covers both process and results with genuine substance. Use this skill at plan-designated checkpoint points — typically after each strategy completes or at key decision nodes within a research Phase. | | context-init | Create a new context file for a research Phase. Called once at Phase start to initialize the file that subsequent context-checkpoint calls will append to. Use this skill whenever a new research Phase begins and a fresh context file is needed. |
<!-- END available-tables (generated) -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | pass→pass | 22,452 | 32,873 | +46% | 1 | 1 | 0% | 3,269 | 5,018 | +54% | 0 | 0 | — |
case-01 | fail→fail | 25,189 | 29,138 | +16% | 1 | 1 | 0% | 3,857 | 5,293 | +37% | 0 | 0 | — |
case-02 | fail→fail | 14,022 | 15,080 | +8% | 1 | 1 | 0% | 2,114 | 2,326 | +10% | 0 | 0 | — |
case-03 | fail→fail | 22,221 | 31,522 | +42% | 1 | 1 | 0% | 3,473 | 4,685 | +35% | 0 | 0 | — |
case-05 | pass→fail | 31,970 | 26,385 | -17% | 1 | 1 | 0% | 5,190 | 5,123 | -1% | 0 | 0 | — |
case-06 | pass→fail | 22,131 | 4,706 | -79% | 1 | 1 | 0% | 3,050 | 1,427 | -53% | 0 | 0 | — |
case-07 | pass→fail | 22,150 | 8,509 | -62% | 1 | 1 | 0% | 3,235 | 1,244 | -62% | 0 | 0 | — |
case-08 | pass→pass | 24,763 | 27,667 | +12% | 1 | 1 | 0% | 3,634 | 4,978 | +37% | 0 | 0 | — |
case-09 | fail→pass | 50,020 | 29,101 | -42% | 1 | 1 | 0% | 1,085 | 5,297 | +388% | 0 | 0 | — |
case-10 | fail→pass | 19,653 | 20,084 | +2% | 1 | 1 | 0% | 3,149 | 3,865 | +23% | 0 | 0 | — |
case-11 | pass→pass | 19,412 | 27,854 | +43% | 1 | 1 | 0% | 2,623 | 4,574 | +74% | 0 | 0 | — |
case-12 | pass→pass | 17,082 | 7,337 | -57% | 1 | 1 | 0% | 2,643 | 1,873 | -29% | 0 | 0 | — |
case-13 | pass→pass | 23,810 | 29,373 | +23% | 1 | 1 | 0% | 3,547 | 5,284 | +49% | 0 | 0 | — |
case-14 | pass→pass | 25,135 | 22,286 | -11% | 1 | 1 | 0% | 3,816 | 4,214 | +10% | 0 | 0 | — |
case-15 | pass→pass | 25,781 | 52,006 | +102% | 1 | 1 | 0% | 4,482 | 4,576 | +2% | 0 | 0 | — |
case-16 | pass→fail | 30,419 | 6,380 | -79% | 1 | 1 | 0% | 5,287 | 1,605 | -70% | 0 | 0 | — |
case-17 | pass→pass | 15,105 | 18,311 | +21% | 1 | 1 | 0% | 2,751 | 3,864 | +40% | 0 | 0 | — |
case-18 | pass→fail | 14,304 | 20,245 | +42% | 1 | 1 | 0% | 2,906 | 4,609 | +59% | 0 | 0 | — |
case-19 | pass→pass | 16,206 | 12,508 | -23% | 1 | 1 | 0% | 3,073 | 3,193 | +4% | 0 | 0 | — |
case-20 | fail→pass | 7,527 | 4,355 | -42% | 1 | 1 | 0% | 1,034 | 1,471 | +42% | 0 | 0 | — |
case-21 | pass→pass | 7,543 | 2,220 | -71% | 1 | 1 | 0% | 1,071 | 1,108 | +3% | 0 | 0 | — |
case-22 | pass→pass | 10,536 | 4,201 | -60% | 1 | 1 | 0% | 1,482 | 1,309 | -12% | 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 20 counted toward the lift figure. The other 2 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 -12 percentage points is the difference between those two pass rates over the 20 comparable cases. 5 cases got worse with the skill loaded, and they are included in that figure.
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.