refactor(skills): simplify current guidance and preserve scan evidence

[drewstone]

Skills repeated full workflows, prescribed arbitrary staffing and measurement thresholds, and copied changing API and model facts.
The revised guidance keeps essential decisions in each entrypoint and loads named supporting procedures only when their conditions apply.
Changing facts point to their current owning source.
Existing consumers still check the package and execution path they actually use.

This covers all 61 dotfiles-owned skills: 59 entrypoints changed; site-clone and session-continuity remain byte-for-byte unchanged.
The related Runtime skill update is tangle-network/agent-runtime#1102.

• Delete 26 competing full-reference documents, the duplicate skill ladder, three redundant shell helpers, and copied scan/loop workflows.
• Remove fixed worker, edit, repetition, score, report, and automation quotas while preserving required behavior and evidence.
• Correct model defaults, research selection criteria, API ownership, release completion checks, and append-only history guidance.
• Preserve security boundaries, durable recovery, experiment identities, comparison integrity, and conditional skill chaining.
• Allow a checked no-change result and short evidence-backed status answers; align the shared defaults with that behavior.
• Fix Semgrep merging to preserve complete runs, zero-finding scans, invocation status, rule indexes, paths, and distinct findings.
  Reject malformed or unsupported input without publishing a partial merge; keep issue deduplication in triage.

Validation:

• RTK_GUARD_MEASUREMENT_CACHE=/tmp/rtk-measurement-aliases-20260904.json npm test: 58 passed, 0 failed, 7 Linux-only skips; smoke passed.
  The unchanged RTK measurement cache was reused; its guard and mutation tests ran.
• 61/61 skill frontmatter validations; final local-reference, logging, and chaining checks passed.
• 7/7 Semgrep behavior regressions passed, including zero findings and preservation of existing output on invalid input.
  These are combiner checks, not complete SARIF schema validation.
• 34 external source links checked against current owning files, trees, or repositories.
• Independent task checks: report identified an incomplete release despite 14 passing tests and HTTP 200; simplify preserved all four files in a sound package.
• Executed the finalize reconstruction, scan filtering, installer preservation, and docs-scanner examples with positive and relevant negative cases.
• Independent source review found one shared reporting contradiction; it was corrected and re-reviewed with no remaining findings.

These checks support instruction integrity and the tested behaviors.
They do not establish a comparative improvement in agent productivity.
Unchanged bundled lint rules were not tested against a fresh Oxlint installation.

Text measurements use whitespace-separated words and physical lines across all 61 owned entrypoints and their supporting Markdown.
Shared author guidance is excluded from these totals.
Baseline is 2178ec6; candidate is 739a096.

Measure	Before	After
Entrypoint words	39,805	25,234
Entrypoint lines	4,483	3,332
Supporting Markdown files	42	39
Supporting Markdown words	42,460	11,426
Entrypoint words: min / median / p90 / max	107 / 550 / 1,318 / 1,671	107 / 410 / 525 / 622

Complete per-skill measurements (before → after)

Skill	Entrypoint words	Entrypoint lines	Supporting files	Supporting words
agent-behavior-audit	330 → 356	42 → 45	1 → 1	797 → 351
agent-eval	483 → 337	58 → 51	0 → 0	0 → 0
agent-integrations-adoption	532 → 364	72 → 53	0 → 0	0 → 0
arena-experiment	462 → 460	82 → 82	0 → 0	0 → 0
autopsy	377 → 373	44 → 46	1 → 1	1181 → 403
breakout	770 → 464	66 → 63	1 → 1	912 → 341
build-agent-app	1364 → 494	188 → 66	0 → 2	0 → 393
calibrate-before-measure	386 → 428	50 → 53	0 → 0	0 → 0
converge	341 → 364	51 → 46	1 → 0	1291 → 0
critical-audit	1318 → 572	119 → 68	6 → 5	3636 → 628
deep-clean	1373 → 454	113 → 56	1 → 0	692 → 0
deploy-proof	756 → 432	96 → 51	0 → 0	0 → 0
diagnose	1290 → 499	120 → 66	1 → 0	512 → 0
director-autopsy	404 → 431	59 → 59	0 → 0	0 → 0
discovery-lead	520 → 468	62 → 58	0 → 1	0 → 290
docs-slop-audit	298 → 335	44 → 48	2 → 1	1146 → 465
dont-collapse-the-architecture	295 → 317	42 → 44	0 → 0	0 → 0
eval-agent	677 → 486	101 → 65	0 → 1	0 → 338
eval-engineering	882 → 558	123 → 72	0 → 1	0 → 255
eval-harness-diagnose	679 → 503	92 → 64	0 → 0	0 → 0
evolve	1454 → 593	121 → 72	4 → 3	5770 → 1393
finalize	869 → 483	60 → 57	1 → 1	2503 → 724
governor	1400 → 622	111 → 73	0 → 0	0 → 0
ground-truth	586 → 490	38 → 59	0 → 0	0 → 0
harden	377 → 399	51 → 49	1 → 1	1852 → 484
harness-escalation-audit	564 → 430	79 → 55	0 → 0	0 → 0
hub-sdk-adoption	619 → 369	82 → 53	0 → 0	0 → 0
hypothesize	829 → 419	59 → 57	1 → 1	863 → 233
install-anti-slop	555 → 341	97 → 48	0 → 1	0 → 108
meta-harness	361 → 460	51 → 59	1 → 1	2024 → 420
model-freshness	598 → 380	86 → 55	0 → 0	0 → 0
multi-pursue	353 → 297	45 → 45	1 → 0	832 → 0
nano-banana	309 → 283	78 → 41	0 → 0	0 → 0
operate	544 → 468	73 → 63	0 → 0	0 → 0
orchestrate	550 → 327	66 → 48	1 → 1	492 → 200
polish	1145 → 386	103 → 51	1 → 1	420 → 253
problem-sourcing	523 → 314	43 → 45	0 → 1	0 → 199
product-design-audit	354 → 410	48 → 51	2 → 1	2797 → 408
product-design	369 → 380	50 → 51	1 → 1	1308 → 242
product-innovation-audit	371 → 429	56 → 50	1 → 1	3111 → 343
pursue	1445 → 594	121 → 70	1 → 0	758 → 0
push-past-easy	486 → 349	43 → 44	0 → 0	0 → 0
reconcile	1006 → 563	97 → 65	0 → 0	0 → 0
reflect	1070 → 525	114 → 65	1 → 2	1160 → 589
refresh-reasoning-capabilities	736 → 316	109 → 46	0 → 0	0 → 0
release-conductor	338 → 384	50 → 48	1 → 1	1061 → 297
report	1671 → 447	148 → 55	0 → 1	0 → 515
review-to-green	688 → 440	57 → 54	0 → 0	0 → 0
sandbox-sdk-integration	335 → 329	46 → 51	1 → 2	469 → 394
semgrep	316 → 444	56 → 62	4 → 3	2360 → 605
session-continuity	359 → 359	53 → 53	0 → 0	0 → 0
ship	315 → 344	50 → 45	1 → 0	889 → 0
signal-distill	292 → 362	44 → 51	1 → 0	868 → 0
simplify	581 → 392	70 → 51	1 → 0	1061 → 0
site-clone	107 → 107	19 → 19	0 → 0	0 → 0
slack-alerts	1057 → 353	72 → 50	0 → 1	0 → 205
tangle-blockchain-blueprint	668 → 396	86 → 55	0 → 0	0 → 0
tangle-ops	874 → 445	71 → 60	0 → 0	0 → 0
ui-test	293 → 336	45 → 44	1 → 1	875 → 350
verify	637 → 416	69 → 54	0 → 0	0 → 0
writing-profile	264 → 358	42 → 52	1 → 0	820 → 0