Event 77 · CP-EXAMPLES-SCHEMA-PARITY-01 · 4 fork-onboarding templates schema-parity rewrite

[junjslee]

Summary

Third v1.0.1 polish mini-batch CP shipped. Brings core/memory/global/examples/ to schema parity with the v2 canonicals so that fork-onboarding via episteme init seeds schema-faithful files instead of stale v1-era templates.

Trigger. Events 65/66/71 privatized the maintainer's canonical profile content; the example templates that fork users seed via episteme init were left at their pre-v2-schema state. The four files in this PR were structurally / schema-wise inadequate to mimic the canonicals — the kernel hooks read canonical paths at runtime and fall back to defaults when axis data is missing, so fork users were getting generic-default kernel behavior instead of personalized-to-them behavior.

What this changes

File	Before	After
operator_profile.example.md	1.6KB / 0 axes (machine-spec content from v1 schema)	8.7KB / 16 axes (full v2 schema scaffolding)
cognitive_profile.example.md	8.2KB / 4 sections	9.0KB / 7 sections (canonical-aligned)
workflow_policy.example.md	2.5KB / 10 sections (structural drift)	3.7KB / 7 sections (canonical-aligned)
agent_feedback.example.md	DOES NOT EXIST	3.8KB / 3 sections (NEW; canonical-aligned)

operator_profile.example.md (full v2 schema rewrite)

Full v2 schema scaffolding: 16 axes with the per-axis metadata structure (value / confidence / last_observed / evidence_refs / note).

• § 4a · Process axes (0–5): planning_strictness, risk_tolerance, testing_rigor, parallelism_preference, documentation_rigor, automation_level (6)
• § 4b · Cognitive-style axes: dominant_lens, noise_signature, abstraction_entry, decision_cadence, explanation_depth, feedback_mode, uncertainty_tolerance, asymmetry_posture, fence_discipline (9)
• § 4c · Expertise map (1; 4 example domain entries showing format)

Each axis defaults to confidence: stub (signals "fork user must elicit") with an inline note field teaching the value range / enum + the kernel mechanism that consumes it (e.g., the preferred_lens_order derived knob reads dominant_lens; the Phase 12 audit floor reads asymmetry_posture).

Includes the canonical's Legacy Working Style Summary section (kept for adapter compatibility) and the Review checklist for the operator (3-question discipline before flipping stub → inferred → elicited).

cognitive_profile.example.md (3 missing sections added; tone reconciled)

Adds 3 sections that matched canonical structure but were absent from the prior template:

• Decision Engine (Operational Thinking Rules) — 8 rules translating philosophy → working cadence
• Cognitive Red Flags — 6 named patterns that trigger slow-down + reframe
• Authoritative Mapping — relationship between this file + workflow_policy.md + project docs

Renames Collaboration Defaults → Collaboration Stance to match canonical. Preserves the existing Foundational Mental Models section (Kahneman / Dalio / Boyd / Munger) which was the strongest part of the prior template.

workflow_policy.example.md (structural reconcile to 7 canonical sections)

Drops v1-era drift sections (Bounded Automation / Cost Acknowledgment / Publication Boundary / Review Gate); cost-acknowledgment subsumed under Risk and Autonomy Policy. New 7-section structure matches canonical:

1. Standard Flow (Frame → Decompose → Execute → Verify → Handoff)
2. Stage Definitions (full per-stage protocol)
3. Signal-over-Noise Rules (5-question pre-action check)
4. Risk and Autonomy Policy (reversibility-tiered autonomy)
5. Project Memory Contract (authoritative-doc hierarchy)
6. Parallelism Policy (worktree/lane discipline)
7. Local Integration (post-edit episteme sync + episteme doctor)

agent_feedback.example.md (NEW)

Creates the missing template for fork users. 3-section canonical structure:

1. How to classify new agent-learned feedback — promotion bar + anti-duplication rules + scope (universal / universal-principled / project-specific) + source (operator-authored vs agent-learned) classification.
2. Universal rules — empty list with HTML-comment format placeholder showing entry shape (rule title / Rule / Why / How to apply).
3. Universal-principled rules — same format, with explicit Triggering pattern field for pattern-gated rules.

Empty rule lists are intentional — fork users accumulate rules from their own operator-agent interactions, not by inheriting the maintainer's. The format placeholders teach the rule shape without leaking maintainer content.

Content discipline

No operator-personal content leaked. Specifically excluded:

• The maintainer's family memory (Positive vs Negative systems anecdote)
• The asymmetry_posture lived-behavior narrative (Events 65/66/67 references)
• Any axis values matching the maintainer's actual elicited values
• The kernel-tone-discipline + no-AI-co-author rules (those are agent-learned during the maintainer's sessions; fork users start with empty rule lists)

What IS shared: the schema structure, the metadata format, the section taxonomy, the Foundational Mental Models commentary (Kahneman / Dalio / Boyd / Munger — these are public intellectual content already cited in kernel/REFERENCES.md).

Verification

• operator_profile.example.md enumerates all 16 v2 axes with valid confidence: stub defaults
• All 4 templates structurally match their canonicals' section count (16 axes / 7 / 7 / 3)
• No personal anecdotes / family memories / lived-behavior narratives leaked
• Cross-references resolve (kernel/OPERATOR_PROFILE_SCHEMA.md path correct from examples/ via ../../../)
• Operator runs episteme init in a fresh project + verifies seeded files match the templates exactly (manual post-merge)

Soak-invariant

ZERO touches to kernel/* / core/hooks/* / core/blueprints/* / src/episteme/* / tests/* / templates/* / labs/*. All edits under core/memory/global/examples/ (template tier).

v1.0.1 polish queue post-Event-77

• ✅ CP-RELEASE-PLEASE-CHKPT-FILTER-01 (Event 75)
• ✅ CP-SYMLINK-RESTORE-01 Part A (Event 76)
• ✅ CP-EXAMPLES-SCHEMA-PARITY-01 (this PR)
• ⏳ CP-SYMLINK-RESTORE-01 Part B (SessionStart hook integration; deferred)
• ⏳ CP-AUDIT-ACK-01 (~1-2 days, episteme profile audit ack CLI)
• ⏳ CP-EXAMPLES-SCHEMA-PARITY-01 Component 5 (optional episteme verify-examples CLI; deferred)

3 of 4 v1.0.1 polish CPs shipped. CP-AUDIT-ACK-01 is the largest remaining piece.

Cross-references

• Spec source: ~/episteme-private/docs/cp-v1.0.1-polish.md § CP-EXAMPLES-SCHEMA-PARITY-01
• Privatization context: Events 65 / 66 / 71 (the canonical privatization that necessitated this work)
• Schema source of truth: kernel/OPERATOR_PROFILE_SCHEMA.md
• Audit trail: ~/episteme-private/docs/PROGRESS.md Event 77 entry (private)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
episteme	Ready	Preview, Comment	Apr 29, 2026 2:40am