docs(research-grounded): add capability-assertion seed pass (dogfood fix)#33
Merged
Merged
Conversation
…fix) Dogfooding the skill across the fleet today, its documented step-2 grep (decision verbs: chose/adopted/instead-of) surfaced 0 of the 4 real gaps found — all four were capability/library assertions carrying no decision verb (kourai's RealtimeTTS word-timing + A2A Message.metadata, phalanx's SQLModel+asyncpg "standard 2026 setup", velocity's built-in 8-bit quant). Add a Pass B grep keyed on capability verbs (supports/exposes/provides/ natively/built-in) + best-practice-as-fact words (standard/canonical/ recommended/modern), and make explicit that proper-noun library/protocol names are the real signal so the design sections must be read directly — neither grep catches a bare "Postgres + SQLModel + asyncpg" list. Checked (2026-05): proselint/write-good/Vale flag style weasel-words, not provenance gaps, so the seed is bespoke by necessity. make test + build green.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Dogfooding
/techne:research-groundedacross the sister fleet this session exposed a real gap in the skill's own workflow.The miss. Step 2's documented grep keys on decision verbs (
chose|switched|adopted|…|instead of). Run against the fleet, it surfaced 0 of the 4 genuine gaps found — and 10+ false positives. All four real gaps were capability / library assertions that carry no decision verb:content_kindonMessage.metadata"The skill already names this class ("Version / API-surface bets stated as fact" — the ElevenLabs-SSML miss), but its seed grep was blind to it. I found all four via capability-signal search + section reading, exactly the manual fallback this change encodes.
The fix. Step 2 becomes two passes + a read-the-sections mandate:
supports/exposes/provides/natively/built-in) + best-practice-as-fact words (standard/canonical/recommended/modern). Catches "X exposes Y", "Standard 2026 setup", "the modern standard is X".Postgres + SQLModel + asyncpglist trips no keyword, so the design sections must be read directly.Checked (2026-05): proselint / write-good / Vale flag style weasel-words ("very", "quite"), not capability-claims-missing-provenance — so the seed is bespoke by necessity, not a reinvented wheel. Recorded as a
research(2026-05)note in the skill.Frontmatter untouched;
make test+ strictmake buildgreen locally.