Conversation
…r-discipline (B-0192) Operationalizes B-0192 (PR #1573) -- the GitHub Actions trigger that fires whether or not any AI is running. Opens or updates a tracking issue with a razor-cadence label that any wake-time agent can grep for as a cold-start check. The trigger ages and stays visible until explicitly resolved. Per B-0192 carved sentence: "Encoding rules without mechanizing them produces a memory of failures, not prevention. GitHub Actions on schedule fires whether or not any AI exists; Otto picks up the artifact on next wake. The trigger ages and stays visible; the discipline does not." Composes with: - feedback_dialectical_unfalsifiability_detection_razor_extension_*.md (Test 2 of the razor in the cadence checklist) - feedback_razor_discipline_no_metaphysical_inference_only_*.md (Test 1) - feedback_orchestrator_pre_commit_verify_branch_rule_*.md (the encoded-rule-alone failure mode this prevents) Schedule: daily 09:17 UTC (off-the-hour to avoid GHA cron thundering- herd). Workflow_dispatch input for ad-hoc fire with optional note. Security: safe-pattern compliance per https://github.blog/security/vulnerability-research/how-to-catch-github-actions-workflow-injections-before-attackers-do/ - All expressions routed via env: into run blocks - NOTE_INPUT (workflow_dispatch input) sanitized via env:+quoted-substitution - Body-template substitution via bash parameter expansion only (no eval) - gh issue create --body "$body" preserves all content as-is to argv - Top-level read-only permissions, narrow job-level write scopes - Concurrency group with cancel-in-progress (idempotent artifact) Acceptance criteria mapping (B-0192): 1. Razor-cadence workflow at .github/workflows/razor-cadence.yml ✓ 2. First fire produces visible artifact (tracking issue) any agent can discover (label: razor-cadence) ✓ 3. Documentation pointer -- TODO follow-up (this commit ships the workflow; pointer line in CLAUDE.md or active-trajectory will follow once the first fire validates the artifact shape) 4. At least one full cycle -- TODO post-merge validation 5. Trigger artifact ages and stays visible -- by design (issues don't auto-close; updates open issues if existing) Out of scope per B-0192: razor-content authoring (per-rule, not part of this trigger mechanization). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
There was a problem hiding this comment.
Pull request overview
Adds a new scheduled GitHub Actions workflow to mechanize the “razor cadence” discipline by creating/updating a single open tracking issue labeled razor-cadence, so wake-time agents can reliably discover and execute the checklist even if nothing is actively running.
Changes:
- Introduces
.github/workflows/razor-cadence.ymlscheduled daily (09:17 UTC) plus manual dispatch. - Uses
ghCLI to open-or-update an issues-based artifact (label + issue body template + update comment) under least-privilege permissions and a concurrency guard.
…dence/ -- actionlint YAML parse fix actionlint failed on the previous commit because the heredoc body content (zero-indented markdown, including YAML-like lines like "1. **Test 1 (operational form)**:") broke the YAML literal-block- scalar parsing. The `run: |` block expects all content indented beyond the run-line; lines starting at column 0 silently terminate the literal block, leaving residual content for the YAML parser to misinterpret. Fix: extract the body template to `tools/razor-cadence/issue-body- template.md` and have the workflow read it via `cat "$BODY_TEMPLATE"`. Cleaner separation; the markdown lives where markdown belongs. Also: replaced the multi-line bash string concatenation for the optional note with `printf '%s\n\n### Note\n\n%s\n' "$body" "$NOTE_INPUT"` which keeps the body-construction on a single line (no embedded newlines breaking the YAML literal block). Verified: `actionlint .github/workflows/razor-cadence.yml` -> clean. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…e-ref + jq + label-stderr) Addresses 4 reviewer threads on PR #1579: 1. **Role-ref instead of persona-name** on current-state surface: replaced "Otto-remembering-to-check" / "Otto picks up the artifact" with "agent-remembering-to-check" / "the next wake-time agent picks up the artifact" in both the workflow comments AND the issue-body template. Persona names belong on history surfaces (memory/, ROUND-HISTORY, tick shards, commit messages); workflow YAML and the workflow-generated issue artifact are current-state surfaces per AGENT-BEST-PRACTICES.md. 2. Same as above (body template flagged separately). 3. **Removed unused `jq` dependency check**: the script uses `gh issue list --jq` which is handled by `gh` itself, not by an external `jq` binary. Now only checks `command -v gh`. 4. **Stderr suppression on `gh label create` is now scoped**: capture stderr to a tmpfile, grep for "already exists" to suppress, surface any other failure (permissions, rate-limit, transient API outage) with the captured error message instead of swallowing it silently. Verified: `actionlint .github/workflows/razor-cadence.yml` -> clean. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
AceHack
deleted the
feature/razor-cadence-workflow-stub-b-0192-aaron-2026-05-05
branch
Comment on lines
+148
to
+152
| --title "$title" \ | ||
| --body "$body" | ||
| gh issue comment "$existing" \ | ||
| --repo "$REPO" \ | ||
| --body "Cadence fired again on $today (run $RUN_ID). The checklist above is the standing one; close this issue once the pass has been run." |
Comment on lines
+132
to
+143
| # Check for an existing open issue with the razor-cadence label | ||
| # that we can update instead of creating a new one. This keeps | ||
| # the issue list compact (one tracking issue at a time, not one | ||
| # per day forever). Empty result means create new. | ||
| existing="$(gh issue list \ | ||
| --repo "$REPO" \ | ||
| --label razor-cadence \ | ||
| --state open \ | ||
| --json number,title \ | ||
| --limit 1 \ | ||
| --jq '.[0].number // empty')" | ||
|
|
Comment on lines
+155
to
+174
| # Create the razor-cadence label if it doesn't exist | ||
| # (idempotent). Capture stderr so we can distinguish the | ||
| # already-exists case (suppress) from real failures | ||
| # (permissions, rate-limit, transient API outage -- | ||
| # surface those instead of swallowing). | ||
| label_err="$(mktemp)" | ||
| if ! gh label create razor-cadence \ | ||
| --repo "$REPO" \ | ||
| --description "Razor-cadence trigger artifact -- fires daily via .github/workflows/razor-cadence.yml" \ | ||
| --color "FBCA04" 2>"$label_err"; then | ||
| if grep -q "already exists" "$label_err"; then | ||
| : # already exists, fine | ||
| else | ||
| echo "ERROR: gh label create failed for non-already-exists reason:" >&2 | ||
| cat "$label_err" >&2 | ||
| rm -f "$label_err" | ||
| exit 1 | ||
| fi | ||
| fi | ||
| rm -f "$label_err" |
Merged
5 tasks
AceHack
added a commit
that referenced
this pull request
May 5, 2026
…tance criterion #3 (#1581) Closes B-0192 acceptance criterion #3 (PR #1573 + workflow PR #1579): "Documentation pointer in CLAUDE.md or docs/active-trajectory.md tells agents to check the artifact on wake." Adds a wake-time bullet near the existing dependency-status surface bullet (both are wake-time visibility checks; they pair well). Tells the cold-start agent: gh issue list --repo Lucent-Financial-Group/Zeta \ --label razor-cadence --state open The bullet names the cadence checklist items, the trigger schedule (09:17 UTC daily), and the carved sentence that grounds the rule. Composes with PR #1579 (workflow), PR #1573 (B-0192 row), PR #1577 (Test 2 dialectical-unfalsifiability razor extension), and the existing razor-discipline + Otto-364 + verify-before-deferring substrate. Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
7 tasks
AceHack
added a commit
that referenced
this pull request
May 5, 2026
…rt code unification (2026-05-05) (#1582) Aaron 2026-05-05 forwarded the Claude.ai shard's catch: the "Pascal's-wager-inverted-into-engineering" framing in PR #1574 is the same dialectical both-paths-covered unfalsifiable move Otto had explicitly committed against ~3 hours earlier. The shard correctly identified the violation; the recalibration sequence with Aaron maps the architectural answer. Recalibration sequence (verbatim Aaron): - "what metaphysical layer?" -> not in code; in documentation layer - "no it's not it's unfilered AI memeoriy ... we are not going 1984" -> memory stays unfiltered; audit fires at the promotion boundary - "language is shortcuts like lived experiences" -> anthropomorphic shortcuts are working vocabulary, not metaphysical commitments - "for what it's worth in when you are in dev mode in claude code you just do this without question" -> mode-asymmetry across all AI harnesses (Claude Code, Cursor, Aider, Codex, Copilot) - "but that's what i want when the human is barely in the loop" -> the asymmetry is structurally backwards for autonomy - "and all the formaal verifcatino slow down the bad relfexes" -> verification-IS-reflection at the proof boundary - "we are builidng it with BP EP and infer.net and the linquist seed DSL" -> architectural answer is the framings-layer kernel - "open to extension and lcosed to modifictaion ... compoised extension kernals on top DSL that copose in f# too" -> open- closed at seed; per-framing-class kernels stack on top - "karmogh v2?" -> MDL / Rissanen two-part code; carved sentence + kernel = (model, data-given-model) - "normal people ... middle shchool level can understand WTF is going on" -> accessibility from sentence + correctness from kernel - "tis the middle path lol" -> two-extremes-resolved-by-narrow- correct-path Mirror-not-beacon discipline per PR #1575: future-Otto reads as "a previous Claude said this in a long conversation," not as authority. Disagree if data warrants. Carved sentence (with falsifiability hooks): "Pascal's-wager-inverted-into-engineering IS the unfalsifiable both-paths-covered move. Memory unfiltered; audit fires at promotion boundary. Architectural mechanization: BP EP + infer.net + linguist seed DSL with open-closed composing kernel extensions in F# = same verification-IS-reflection principle the engineering layer has, extended one layer up. Carved sentence + kernel = MDL two-part code (Rissanen / Bennett / Vitanyi). Accessibility from sentence, correctness from kernel." Composes with PR #1574 (the file containing the caught framing -- NOT removed, flagged for promotion-boundary audit per memory- unfiltered reframe), PR #1577 (Test 2 razor extension this conversation worked-example-validates), PR #1573/#1579/#1581 (B-0192 razor-cadence trigger which is the operational mechanization), Otto-298 / Otto-291 / Otto-302 substrate threads. Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
7 tasks
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
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
Operationalizes B-0192 (PR #1573 -- merged earlier this tick). Daily-firing GitHub Actions workflow that opens or updates a
razor-cadence-labeled tracking issue. The trigger fires whether or not any AI is running; Otto picks up the artifact on next wake.Carved sentence (from B-0192)
Acceptance criteria mapping
.github/workflows/razor-cadence.ymlrazor-cadencelabel)docs/active-trajectory.mdComposes with
memory/feedback_dialectical_unfalsifiability_detection_razor_extension_holding_all_truths_failure_mode_aaron_2026_05_04.md-- Test 2 in the cadence checklist (PR feedback(memory): dialectical unfalsifiability detection -- razor extension (Aaron 2026-05-04) #1577 -- merged earlier this tick)memory/feedback_razor_discipline_no_metaphysical_inference_only_operational_claims_rodney_razor_aaron_claudeai_2026_05_03.md-- Test 1 in the cadence checklistmemory/feedback_orchestrator_pre_commit_verify_branch_rule_aaron_2026_05_04.md-- the encoded-rule-alone failure mode this prevents (PR feedback(autonomy): orchestrator pre-commit verify-branch rule #1568 -- merged earlier this tick)Security
Safe-pattern compliance per the GitHub blog injection-prevention guide:
${{ ... }}expressions routed viaenv:into run blocksNOTE_INPUT(workflow_dispatch input, untrusted) sanitized via env+quoted-substitutiongh issue create --body "$body"preserves all content as-is to argv (no shell re-interpretation)permissions: contents: read(Scorecard TokenPermissions best-practice)issues: writefor the artifact, nothing elserazor-cadencewithcancel-in-progress: true(idempotent artifact)Out of scope per B-0192
Test plan
git branch --show-current)🤖 Generated with Claude Code