consolidate: curate learnings, skills, and guidelines (2026-02-28) - #13
Merged
Conversation
29 skills across 5 namespaces evaluated. Fixed ambiguous reference path in do-refactor-code (bare filename → full ~/.claude/ path). 1 LOW recorded: orphaned subagent-patterns.md in skill-references. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
B-1: Fold validation.md (unreferenced guideline) into learnings/skill-design.md L-1: Delete orphaned skill-references/subagent-patterns.md, remove stale reference from multi-agent-patterns.md Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
5 learnings compounded from consolidation loop into skill-design.md and ralph-loop.md. Remove consolidate-output/ working files from PR diff. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Instead of deleting the orphaned reference file, wire it into the 3 skills that launch subagents (explore-repo, do-security-audit, parallel-plan:execute). Restores cross-reference from multi-agent-patterns.md. Removes the now-incorrect "orphaned skill-references" learning from skill-design.md. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…anup rule Extend MEDIUM auto-apply reference wiring to cover skill-reference files (not just personas). Add tiebreaker principle: when a finding has both a cleanup option (delete) and an opportunity option (wire/restructure), prefer opportunity — it fixes root cause and is reversible. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Consolidation agent now persists meta-insights directly into the learnings system during the loop instead of staging them to compounded-learnings.md for manual post-loop compounding. Removes compounded-learnings.md from output files — compound actions log to decisions.md like all other actions. Also: session-start learnings gate now uses ambient context (branch, CWD, git status) with filename-only glob for cheaper compliance, and two new learnings on Skill tool limitations in worktrees. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ahoym
added a commit
that referenced
this pull request
Mar 13, 2026
Learnings extracted from algo-trading PRs #3-#13, covering: - Code quality: dead code removal, guard variables, dict inlining, function reuse - Code review: exceptions vs Optional, accurate summaries, feature naming - Python: __post_init__ for derived dataclass fields - Git: stacked PR risks, dependency chain abandonment - Process conventions (new file): pyproject.toml anchoring, migration scripts, cross-cutting refactor deferral, plan-first PRs Also updates extract-request-learnings skill templates to strip Source/Frequency from general learnings (learning matters, not provenance) and only show Frequency on project learnings when value is "convention". Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
7 tasks
ahoym
added a commit
that referenced
this pull request
May 17, 2026
Addresses team review (#12, #13, #14) on PR #109. - `cutoff = time.time() - args.days * 86400` (wall-clock) instead of anchoring to the newest log entry. `--days 7` now means "last 7 calendar days" rather than "7 days before last activity". - `import argparse` and `import time` move to the top-level import block. - All four percentage calculations use `round(100 * n / d)` instead of integer floor division: overall suggestion-fire rate, top-performers rate, and both tier hit rates. 2-of-3 now reads as 67%, not 66%. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ahoym
added a commit
that referenced
this pull request
May 18, 2026
Addresses team review (#12, #13, #14) on PR #109. - `cutoff = time.time() - args.days * 86400` (wall-clock) instead of anchoring to the newest log entry. `--days 7` now means "last 7 calendar days" rather than "7 days before last activity". - `import argparse` and `import time` move to the top-level import block. - All four percentage calculations use `round(100 * n / d)` instead of integer floor division: overall suggestion-fire rate, top-performers rate, and both tier hit rates. 2-of-3 now reads as 67%, not 66%. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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.
Summary
Automated consolidation sweep of
~/.claude/knowledge corpus — learnings, skills, and guidelines. Converged after 3 rounds (9 sweeps) with 4 HIGHs, 3 MEDIUMs applied, 1 MEDIUM blocked.Key actions:
research-methodology.md(content already inskill-design.md) andparallel-planning.md(merged intoparallel-plans.md)parallel-plans.mdalready covered inclaude-code.mdxrpl-testing-patterns.md(1 pattern, 16 lines) intoxrpl-patterns.mdxrpl-typescript-fullstackpersona with Detailed references to XRPL learningsdo-refactor-code/SKILL.mdto use full~/.claude/learnings/pathsquantum-tunnel-claudes/SKILL.mdBlocked (needs human decision):
guidelines/validation.mdis not@-referenced inCLAUDE.md— whether to wire it as always-on is a preference callChanges
Learnings (35 → 32 files):
research-methodology.md,parallel-planning.md,xrpl-testing-patterns.mdparallel-plans.md(deduped sections),xrpl-patterns.md(absorbed testing patterns)xrpl-patterns.mdentries from folded fileSkills:
do-refactor-code/SKILL.md— fixed 2 bare filename refs → full~/.claude/learnings/pathsquantum-tunnel-claudes/SKILL.md— fixed reference pathPersonas:
xrpl-typescript-fullstack.md— added Detailed references section (4 learnings files linked)Consolidation output (new, metadata only):
consolidate-output/— spec, progress, report, decisions, blockers, lows, compounded-learningsScreenshots
N/A
Test plan
@-references in CLAUDE.md after mergeconsolidate-output/blockers.md— decide onvalidation.mdwiringChecklist
compounded-learnings.mdfor insights to persist via/learnings:compoundvalidation.mdblockerGenerated with Claude Code