canonical-doc-contract-plan-hardening
Harden a documentation or mission-contract plan so adversarial reviewers cannot reject it for vague validator contracts, wrong artifact authority, or semantic drift gaps.
Best use case
canonical-doc-contract-plan-hardening is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Harden a documentation or mission-contract plan so adversarial reviewers cannot reject it for vague validator contracts, wrong artifact authority, or semantic drift gaps.
Teams using canonical-doc-contract-plan-hardening should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/canonical-doc-contract-plan-hardening/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How canonical-doc-contract-plan-hardening Compares
| Feature / Agent | canonical-doc-contract-plan-hardening | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Harden a documentation or mission-contract plan so adversarial reviewers cannot reject it for vague validator contracts, wrong artifact authority, or semantic drift gaps.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# Canonical doc-contract plan hardening Use when drafting or revising a plan for a canonical documentation contract: repo mission, role map, glossary, or cross-doc wording reconciliation. ## Common reviewer failures this prevents 1. Validator contract is too vague - Reviewers reject plans that say "add a validation script" without listing exact required phrases, forbidden phrases, section expectations, or file-specific rules. 2. Canonical artifact stored in the wrong namespace - Do not put a normative long-lived contract under a reporting/history path unless that repo explicitly treats the path as authoritative. 3. Phrase-only validation is too weak - Required/forbidden strings alone are insufficient. Add semantic role-claim checks and file-specific expectations. 4. Artifact status is muddy - Split artifacts into: - existing evidence inputs - planned outputs - Do not mix prior review artifacts with future deliverables in the same ambiguous list. 5. Cross-document terminology drift remains unresolved - Explicitly distinguish concepts that share similar words. Example: - ecosystem control plane - workflow control plane inside that ecosystem ## Required hardening pattern ### 1. Put the canonical artifact in a normative location Prefer a path under a durable standards/contracts namespace or another clearly authoritative docs path. Avoid using report/history namespaces for the canonical artifact unless the repo already treats them as authoritative. ### 2. Add a "Canonical Terminology Contract" section to the plan This section should include all of the following: - exact required phrases for the canonical contract doc - exact forbidden phrases - required non-goal bullets - required glossary terms when the deliverable claims to include a glossary - file-specific expectations for each reconciled document - explicit validator semantics: - case sensitivity - line-ending normalization - whitespace handling - whether checks are whole-line, substring, or regex-based - exact plan-index rule if the plan updates a local index table, including full table schema when relevant Also add explicit migration checks for legacy wording that must be removed, not merely supplemented. Example: if an existing doc says `GSD is the control plane`, add that exact legacy phrase to the forbidden list once the new distinction (`workspace-hub` as ecosystem control plane vs `GSD` as workflow control plane) is introduced. ### 3. Split lexical and semantic checks Your TDD/validation plan should include both: - lexical checks - phrase presence/absence - required section headers - required bullets - glossary terms when claimed in the deliverable - semantic checks - no non-canonical repo gets assigned the canonical role - role statements in secondary docs do not contradict the main contract - generic standards docs stay generic and do not absorb repo-specific ownership prose - legacy role phrasing is removed where required, not merely coexisting beside the new phrasing - related standards/contracts cross-link to each other when the plan introduces two nearby authoritative docs ### 4. Lock constrained entrypoint files explicitly If a root workflow-contract file is constrained by line count or by role, measure that in Resource Intelligence and make the plan explicit: - unchanged in this packet - or changed only under a deterministic rule Also capture a deterministic baseline when immutability matters: - current line count (for example via `wc -l`) - current git blob SHA when possible (`git rev-parse HEAD:path/to/file`) Use that baseline in the TDD plan so a later `file unchanged` test has a concrete source of truth rather than a vague before/after comparison. Do not leave edits to a constrained file as an "optional if needed" judgment call. ### 5. Separate existing review evidence from future outputs Artifact Map should usually have two subsections: #### Existing evidence inputs - this plan - steering docs - standards docs - already generated review artifacts #### Planned outputs - canonical contract doc - validation script - reconciled target docs - follow-up issue drafts ### 6. If manual validation would rot, draft the CI follow-up now If the current issue is only for creating the contract plus validator, add a follow-up issue draft for CI enforcement so reviewers can see drift prevention is queued. Be explicit about whether the follow-up draft is being created or updated. If a draft file already exists, treat it as a modify action, not create, and make the acceptance criterion require concrete contents: - validator script path - test file path - exact test command - intended CI hook/job ### 7. Review-wave hygiene for iterative plan hardening When a plan goes through multiple adversarial-review waves: - keep the review-artifact header in sync with the actual generated files - keep the artifact map explicit about which waves already exist versus which outputs belong to eventual implementation - do not leave the plan self-reporting `FAIL` once you have patched the cited blockers; change the summary to something like `pending final delta review` and immediately rerun review - convert lingering open questions into explicit decisions or explicit deferred follow-up items before asking for approval - if reviewers keep objecting to semantics rather than mechanism, add the exact regexes, section shapes, or file-specific predicates to the plan itself ### 8. Standards-vs-mission boundary rule If a new canonical mission contract sits near an existing generic standard: - keep the mission contract in a normative location - keep the generic standard generic - if cross-links are required, specify the exact relative markdown links and the relationship note expected in each file - do not make a generic cross-repo standard carry repo-specific ownership tables just to satisfy a validator - most importantly: do not leave the standard's edit status ambiguous. If tests/acceptance require a cross-link or wording change, mark the file as a mandatory generic-only modify everywhere in the plan. Do not call it optional in one section and required in another. ### 9. Validator semantics must be concrete, not inferred When reviewers keep attacking validator ambiguity, add the exact semantics to the plan itself: - how fenced code blocks are detected (for example, triple-backtick fences only) - whether indented code blocks are exempt or not - whether checks exclude fenced content consistently across required-phrase, forbidden-phrase, and semantic-regex rules - whether matching is case-sensitive - whether line endings are normalized - whether Unicode is normalized (for example NFC) - whether paragraph line-wraps are collapsed to spaces before matching - whether a forbidden rule is substring-based or a whole-line regex If a legacy phrase is being retired, prefer an explicit standalone regex over a vague substring ban. ### 10. Immutable file checks need a reproducible baseline If the plan says a constrained file must remain unchanged, the test should not rely on a vague before/after claim. Capture and use a deterministic baseline such as: - `git rev-parse HEAD:path/to/file` for a blob SHA - current line count via `wc -l` Then make the test assert the working-tree blob still matches the recorded SHA when immutability is part of the contract. ### 11. Keep review-wave bookkeeping synchronized In multi-wave hardening loops: - if review artifacts are listed in plan metadata, keep the list synchronized with every actual review wave you cite in the summary - if the Adversarial Review Summary mentions wave N, the artifact map and metadata should also reflect wave N - if you patch blockers from the latest wave, change the summary status from outright FAIL to a pending-final-delta state only after you have actually made the patches and immediately rerun review - do not let the plan's self-reported status lag behind the current draft ### 12. Align prose rules with tests verbatim Common reviewer complaint: the prose says one thing and the test row says another. Before re-review, cross-check all of these for exact agreement: - Files to Change - Artifact Map - Canonical Terminology Contract - TDD Test List - Acceptance Criteria - Adversarial Review Summary revision notes Typical drift to catch: - optional vs mandatory edits - partial vs full plan-index schema - create vs modify on already-existing follow-up drafts - semantic checks applying to reconciled docs but not the canonical contract itself ## Template additions that help approval Include concrete file-specific expectations like: - a given top-level doc must contain the canonical role statement - a strategy/brain doc must preserve its stronger structure while normalizing wording - a generic standard may only receive generic terminology edits and must not gain repo-specific role tables Include concrete non-goals like: - repo does not own engineering computations - repo does not own shared utility layer - repo does not own GTM/public website - repo does not resolve an open architecture decision tracked in another issue ## Example validation dimensions For a mission/control-plane contract validator, check all of: - required phrases in canonical contract - forbidden phrases across reconciled docs - required non-goal bullets in canonical contract - file-specific role statements - semantic no-contradiction checks for named repos - generic standard remains generic - local plan index row present with expected schema - follow-up CI issue draft exists if validator is not yet enforced automatically ## Practical rule If a reviewer says "the mechanism is specified but the substance is not," add more explicit content to the plan itself rather than only improving the future validator script. ## Outcome to aim for A plan becomes approval-ready when: - canonical location is correct - content contract is explicit in the plan - validator behavior is mechanically derived from that contract - semantic contradictions are blocked, not just missing phrases - artifact status is unambiguous - any deferred enforcement work is captured in a follow-up issue draft
Related Skills
llm-wiki-page-shape-contract
Enforce the page-shape contract when a repo-side document or analysis output gets converted into an llm-wiki page. Use when (1) running `scripts/knowledge/llm_wiki.py ingest`, (2) writing or rewriting a wiki page from docs/reports/*, docs/handoffs/*, scripts/review/results/*, or calc citation outputs, (3) deciding whether a page should be split into a folder of sub-pages, (4) reviewing wiki PRs for length / diagram / divide-and-conquer compliance. Codifies the Karpathy + Astro-Han + lewislulu page rules applied to workspace-hub's domain-wiki layout under /mnt/local-analysis/llm-wiki/wikis/<domain>/. Sibling to research/llm-wiki (which owns the CLI ops) — this skill is the quality gate every converted page must clear before commit.
blender-worktree-test-hardening
Recover and harden digitalmodel Blender automation work in isolated worktrees when uv/editable dependency paths break and local machines lack a Blender executable.
plan-review-prompt-refresh-after-plan-edits
Refresh reviewer prompt files from the latest on-disk plan before every adversarial re-review. Prevents Codex/Gemini from critiquing stale plan text after local edits.
workspace-knowledge-doc-contracts
Class-level workspace knowledge, LLM-wiki, repo mission contracts, stale doc references, semantic taxonomy, and knowledge-source reconnaissance.
workspace-hub-overnight-plan-monitor
Monitor and reconcile workspace-hub overnight planning or implementation batches, including process status, result artifacts, issue/commit verification, and controlled failed-lane recovery.
plan-gated-issue-execution-wave
Execute a multi-issue architecture/planning wave in a plan-gated repo, then safely transition approved issues into implementation with file-based Codex prompts, local approval markers, subprocess monitoring, and cleanup handling for sandbox/hook edge cases.
mixed-ops-vs-repo-fix-plan-boundary
Plan mixed operational-vs-repo remediation issues by proving live-state classification first, then only proposing code changes for confirmed repo-owned failure paths.
wave-based-parallel-plan-execution
Orchestrate phase execution by discovering dependencies, grouping into waves, spawning subagents, and collecting results with optional wave filtering
plan-gated-issue-validation-workflow
Systematic validation pattern for plan-approved GitHub issues with pre-existing deliverables
plan-gated-issue-implementation
Workflow for executing pre-approved GitHub issues with mandatory validation checkpoints
multi-role-agent-contract-review-pipeline
Execute a 4-role agent team (Planner/Architect/Reviewer/Integrator) pipeline for self-reviewing knowledge artifacts before delivery
external-drive-ingest-planning
Plan safe external-drive ingests into repo-aligned storage such as /mnt/ace: read-only mounts, manifests, staged rsync, dedupe-merge gates, GitHub issue traceability, and governance/execution split.