conflation-score
Skill49 starsChanged 14 days agoApache-2.0View on GitHub
One repeatable pass that keeps every number and status fak reports PROVENANCE-HONEST - each value labeled by what fak CONTROLS (witnessed/authored) vs what it only OBSERVES (relayed from an external party), and no bad observed value blamed on a fak action. Runs the conflation scorecard (`fak conflation-scorecard`, the Go verb backed by internal/conflationscore) over the fact-reporting surfaces (Prometheus metric help, the fak guard exit summary), turns each HARD defect into a required edit (label an unlabeled external value OBSERVED; correct. Use when this named workflow matches the task.
In this repository151 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- llms-full.txtllms.txt
- llms.txtllms.txt
- agent-readinessSkill
- appeal-scoreSkill
- bottleneck-mapSkill
- ci-repairSkill
- claim-repro-scoreSkill
- clear-out-wipSkill
- commit-cleanSkill
- cross-validateSkill
- curate-clusterSkill
- debt-cleanSkill
- debt-orchestratorSkill
- disambiguation-scoreSkill
- dojo-rsi-scoreSkill
- dos-dispatch-loopSkill
- dos-dispatchSkill
- dos-next-upSkill
- dos-replanSkill
- dos-witness-claimSkill
- field-borrowSkill
- fleet-waveSkill
- git-subagent-syncSkill
- goalSkill
- guard-rsi-scoreSkill
- harness-creatorSkill
- harness-gardenSkill
- industry-scoreSkill
- issue-orchestratorSkill
- issue-queueSkill
- issue-triageSkill
- lightgap-scoreSkill
- memory-compactSkill
- milestone-scoreSkill
- modularizeSkill
- negframe-scoreSkill
- opencode-waveSkill
- operator-heaviness-scoreSkill
- persona-scoreSkill
- phased-planSkill
- plan-auditSkill
- quality-scoreSkill
- question-loopSkill
- refresh-cachedoc-numbersSkill
- refresh-readmeSkill
- releaseSkill
- resume-watchdog-auditSkill
- run-it-all-nightSkill
- score-2xSkill
- scorecardSkill
- scout-loopSkill
- signal-firstSkill
- skill-lifecycleSkill
- skill-overlapSkill
- skill-scoreSkill
- slop-scoreSkill
- sota-checkSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Distinctive: line 30
provenance_labeled (HARD): - every help/summary string that reports an EXTERNAL value (a provider counter, upstream-reported usage) carries …
- Distinctive: line 33
no_false_attribution (HARD): - no prose attributes a bad observed value to a fak action ("the cache broke", "the splice is producing", "we …
- Distinctive: line 50
Weigh the SOFT signal: - if a mixed family names no fault signal, add the one sentence.
- Very specific: line 54
Commit only the scorecard lane, by explicit path: - run fak sync reconcile --apply, then commit via fak commit --path (never git add -A), …
- Very specific: line 60
Commit the scorecard source/test (internal/conflationscore/), the reporting-surface file(s) whose prose you corrected, and the control-pane …
- Very specific: line 75
When a new file renders operator-facing numbers (a metric family, a new exit summary), add its path to ReportingSurfaces in …
- Very specific: line 81
Keep the scorecard source (internal/conflationscore/) and the surfaces it grades ASCII-only (no em-dashes / smart quotes).
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
conflation-score - the anti-conflation / provenance-honesty pass
An instance of the generic scorecard doctrine. Read that for the five laws and the RSI loop; this file is the per-surface specialization.
Why this scorecard exists
The lesson it generalizes (from the compaction-metrics fix): a system that reports numbers
must keep a hard line between what it controls and what it only observes. fak
guarantees the prefix it ships is byte-identical (witnessed, in-process); whether the
provider reuses the cache is the provider's call (observed, relayed). The original metric
prose said a cratered cache_read meant "the cache broke" - blaming fak for a provider-side
miss (TTL expiry, eviction, the client moving its own breakpoint). That is a
truth-maintenance bug: it erodes trust in the one number fak can actually stand behind.
The anti-conflation law, in one line: label every reported fact by its provenance (WITNESSED vs OBSERVED), and never attribute a bad observed value to a fak action unless a witnessed signal proves the fault is fak's.
The KPIs (what it grades)
- provenance_labeled (HARD) - every help/summary string that reports an EXTERNAL value (a provider counter, upstream-reported usage) carries an OBSERVED/provider-relayed qualifier. One unlabeled external value = one debt.
- no_false_attribution (HARD) - no prose attributes a bad observed value to a fak action ("the cache broke", "the splice is producing", "we re-bill") without a co-located disambiguating qualifier. One violation = one debt.
- fault_signal_isolated (SOFT) - a family mixing witnessed + observed values names
exactly one fak-fault signal the reader can point at (the "only
prefix_mismatch>0 is our bug" pattern).
The pass (the shared five-step loop)
- Run it -
fak conflation-scorecard(work-list),--json(payload),--compare baseline.json(prove the drop),--markdown(regenerate the committed snapshot). The card is a Go verb now (internal/conflationscore); the oldpython tools/conflation_scorecard.pywas removed when the card was ported to Go. - Retire conflation_debt worst-first - fix the heaviest KPI by editing the prose: add the OBSERVED label to an external value, or rewrite an attribution sentence so it names the provider-side cause and the single fak-fault signal. This pass changes only comments and help/summary STRINGS - never a number, a counter, or a branch.
- Weigh the SOFT signal - if a mixed family names no fault signal, add the one sentence; don't invent signals that don't exist.
- Re-measure + prove -
--compareprints the debt delta; the scorecard must read A (debt 0) on the disciplined tree. - Commit only the scorecard lane, by explicit path - run
fak sync reconcile --apply, then commit viafak commit --path(nevergit add -A), and push viafak sync push:
Commit the scorecard source/test (fak sync reconcile --apply fak commit --path internal/conflationscore/ --path <reporting-surface-files> -m "docs(conflation): label observed values (fak conflation)" fak sync pushinternal/conflationscore/), the reporting-surface file(s) whose prose you corrected, and the control-pane baseline. End the subject with(fak <leaf>).
The anti-gaming rule (specific to this surface)
Fix a conflation by making the prose honest, never by deleting the number or muting the
detector. A flagged external value is fixed by labeling it OBSERVED, not by removing the
metric. A false-attribution sentence is fixed by naming the real provider-side cause, not by
deleting the sentence so the substring no longer matches. If a flagged string is already
honest in different words (e.g. "provider-side reuse, distinct from the local caches"), the
fix is to teach the detector that phrasing in observedQualifiers
(internal/conflationscore/conflationscore.go) - recognizing real honesty the keyword list
was blind to is NOT weakening the check; relaxing it to pass a genuine conflation is.
Adding a surface
When a new file renders operator-facing numbers (a metric family, a new exit summary), add
its path to ReportingSurfaces in internal/conflationscore/conflationscore.go. The scorecard
then holds it to the same provenance discipline, and the live-tree test pins the new floor.
ASCII gotcha
Keep the scorecard source (internal/conflationscore/) and the surfaces it grades ASCII-only
(no em-dashes / smart quotes). The repo's provenance pre-commit hook crashes on non-cp1252
UTF-8 before its own escape fires. Use - for a dash and straight quotes.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

