Repository navigation
Fix SESSION_RUNNER 3C: route project learnings to CLAUDE.md, reserve SESSION_RUNNER table for framework learnings - #25
Merged
Conversation
…ESSION_RUNNER table for framework learnings
Phase 3C ("Document Learnings") told sessions to "Update the workstream
document and/or the Learnings table below" — pointing adopters at
SESSION_RUNNER's own table. That contradicts BOOTSTRAP.md, which marks
SESSION_RUNNER as a synced file that must stay byte-identical ("Customizations
Go in CLAUDE.md, Not in Synced Files") and directs project learnings to
CLAUDE.md -> Project-Specific Methodology Adaptations -> Project-specific
Learnings. Flagged as Finding KJ5HST#10 in the rad-con runner upstream audit.
Fix routes learnings by audience and removes the contradiction in all three
places it lived:
- SESSION_RUNNER 3C: recomposed to branch on audience — adopter projects
record learnings in CLAUDE.md (and are told NOT to edit the synced table);
the canonical methodology repo, which has no CLAUDE.md Adaptations section,
appends framework learnings to the SESSION_RUNNER "Learnings (added by
sessions)" table (append-only).
- SESSION_RUNNER Learnings-table caption: replaced the false "This table
starts empty" line (it ships with 6 seed rows) with the same dual-audience
framing, so the contradiction does not resurface 120 lines below.
- HOW_TO_USE.md 3C summary: retargeted to match (kept the 3C label per that
block's 3A-3F convention) so it does not become a stale destination.
Canonical vocabulary reused verbatim from BOOTSTRAP.md and CLAUDE_TEMPLATE.md;
the six learning-content bullets and the closing goal line are unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This was referenced Jun 6, 2026
KJ5HST
pushed a commit
that referenced
this pull request
Jun 12, 2026
…o PROJECT_LEARNINGS.md when it grows The Project-Specific Methodology Adaptations → Learnings table accumulates every session with no cap, but CLAUDE.md is loaded in full each session and has a practical size budget (Claude Code targets ~200 lines; an oversized memory file measurably degrades instruction adherence). Teach the overflow valve: when CLAUDE.md nears its budget, extract the growing subsection to a committed, referenced PROJECT_LEARNINGS.md and leave a read-on-demand pointer — the same discipline as the three-file BACKLOG/CHANGELOG/ROADMAP split. Edits are confined to project-owned, non-synced templates (CLAUDE_TEMPLATE.md, BOOTSTRAP.md). The synced trio (SESSION_RUNNER.md, SAFEGUARDS.md, methodology_dashboard.py) is untouched, so adopter syncs are unaffected. Companion to #24/#25, which route project learnings INTO CLAUDE.md; this bounds what happens when that table overflows. No principle, phase, gate, or workstream changes. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
KJ5HST
pushed a commit
that referenced
this pull request
Jun 12, 2026
rmsharp
added a commit
to rmsharp/methodology
that referenced
this pull request
Jun 12, 2026
…n with 2026-06-12 execution record Plan of record drafted 2026-06-06; executed 2026-06-12 after both PRs merged verbatim into KJ5HST/main. Execution record table documents what each adopter project already had, what was added, and where it landed (wsfct#520, mts#1, airqinodashboard#1, nprcgenekeepr local branch). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
KJ5HST
pushed a commit
that referenced
this pull request
Jun 20, 2026
…+ Phase 3F hook — issue #19 The v2.6 cycle produced three instances of one self-review failure mode that git-diff-only review cannot catch — the defect is an *absence* on the destination side: - "24 failure modes" stale-by-one after FM #25 appended on a sibling file - /diagnose and /grill-with-docs index rows understated their citation sites - /init aspirational citation: index claimed a destination that didn't cite back Formalize the heuristic (issue #19, placement option c — both an authoritative Learning row and a workflow hook): - Learning #7 "Cross-reference completeness at self-review" appended to the canonical "Learnings (added by sessions)" table (append-only). - One-line trigger in Phase 3F (Commit), alongside the existing debug-instrumentation removal gate, citing Learning #7. Reconciliation with the issue: option (a) named ITERATIVE_METHODOLOGY.md §Knowledge Accumulation, but that section only *describes* the four accumulation mechanisms and §Session Learnings is a per-session close-out template — neither holds named-Learning rows. The actual canonical Learnings table is in SESSION_RUNNER.md, and Phase 3C routes canonical-repo framework learnings there. Both edits therefore land in SESSION_RUNNER.md (one file). Self-reviewed with the discipline it documents: the "(Learning #7)" reference resolves to the new row; growing the table 6->7 drifts no current-count claim (the only count mentions are dated changelog deltas and frozen docs/planning audits, verbatim by design). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This was referenced May 26, 2026
rmsharp
added a commit
that referenced
this pull request
Jun 23, 2026
Several non-dated current-state sections had drifted from the actual repo: - "three layers" -> "four layers" (campaign templates became a first-class layer in v2.4; the hierarchy table + "All four are needed" already say four) - "23 known failure modes" -> 26 (FM #24 v2.3, #25 v2.6, #26 v2.7) - degradation detection "7 warning signs" -> 16 (the table grew with the FMs) - "What's in the starter kit" table + Repository Structure tree: add the two missing starter-kit files CONTEXT_TEMPLATE.md and RECOMMENDED_SKILLS.md (both shipped v2.6, already named in the Quick Start corpus list) - Repository Structure bin/ block: add check-links, _manifest.py, tests.sh (added in v2.8; the tree still listed only sync + status) Dated "What's New in vX" entries are left verbatim by design. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
rmsharp
added a commit
to rmsharp/methodology
that referenced
this pull request
Jun 25, 2026
Track the three open adopter PR KJ5HST#25/KJ5HST#27 remediations (wsfct #520, mts #1, airqino #1) plus methodology-repo housekeeping as backlog entries. Each adopter entry captures the two operator constraints: update to v2.9 before merge, and merge only when that repo is between sessions. Backlog, not GitHub issues, by operator decision. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
rmsharp
added a commit
to rmsharp/methodology
that referenced
this pull request
Aug 16, 2026
…lation can be a confound From the BL-36 audit: when a defect report names a cause from a correlation that holds across every artifact in existence, ask what else is collinear with it and construct the missing cell rather than reading the evidence harder. Companion half: a failing proof is evidence about the proof until its claim is re-derived by machinery it does not share. Appended at the end of the table; KJ5HST#14 stays reserved. Distributed file, so the row cites artifacts in backticks and adds no repo-relative link. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
rmsharp
added a commit
to rmsharp/methodology
that referenced
this pull request
Oct 6, 2026
Receipt S255 complete (handoff 3A of S254 9/10, self 7/10), the ledger entry that discharges the claim's CHANGELOG: pending, and fork learning #106 (an arm contrast is a property of each saved set's start commit). No row retired: none qualifies (rows considered: KJ5HST#25, KJ5HST#65, #105). Gates: 12/12, suite 362 passed / 0 failed. Co-Authored-By: Claude Sonnet 5.5 <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
2 participants
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.
Closes #24.
What
Phase 3C "Document Learnings" told sessions to "Update the workstream document and/or the Learnings table below" — pointing adopters at SESSION_RUNNER's own table. That contradicts
BOOTSTRAP.md, which marksSESSION_RUNNER.mdas a synced, byte-identical file ("Customizations Go in CLAUDE.md, Not in Synced Files") and routes project learnings toCLAUDE.md→ Project-Specific Methodology Adaptations → Project-specific Learnings. See #24 for the full write-up (originally rad-con audit Finding #10).The fix — three edits, no stale destination left behind
SESSION_RUNNER.md3C — recomposed to route by audience:CLAUDE.md→ Project-Specific Methodology Adaptations → Project-specific Learnings; do NOT edit the synced "Learnings (added by sessions)" table.CLAUDE.mdAdaptations section) → append framework learnings to the SESSION_RUNNER seed table (append-only).SESSION_RUNNER.mdLearnings-table caption — replaced the false "This table starts empty…" line (it ships with 6 seed rows) with the same dual-audience framing, so the contradiction doesn't resurface ~120 lines below.HOW_TO_USE.md:763— retargeted the 3C summary to match (kept the3Clabel per that block's 3A–3F convention) so it isn't a stale destination.Notes
BOOTSTRAP.md(Step 5 heading, "synced from canonical / byte-identical / block future syncs", "Agents read CLAUDE.md at session start") andCLAUDE_TEMPLATE.md("Project-Specific Methodology Adaptations", "Project-specific Learnings"). No invented names; destinations cited by name, not step number.upstream/main; the diff is exactly these three edits.🤖 Generated with Claude Code