Post-merge cleanup: whitespace, doc sync, canonical skill names, root CLAUDE.md (#33)

Cleanup pass following the CoCounsel merge (#4) and Lexis removal (#5):
- Normalize whitespace and JSON indentation across plugin metadata files
- Sync marketplace.json with plugin.json (ip-legal description, Courtroom5 title field)
- Add missing ai-inventory and invention-intake rows to README skill reference
- Replace stale short-form skill names in customize/cold-start SKILL.md prose with canonical directory names across 11 plugins (same bug class as the /setup -> /cold-start-interview QA fix)
- Fix FYY -> FYI typo and a phantom WebFetch claim in launch-radar docs
- Address review findings: line-wrapped /check-claims, /renewals-due -> /renewal-tracker, stale /setup in references/, agent.yaml comment contradiction, redundant 'intake intake'
- Add root CLAUDE.md with validation conventions (claude plugin validate, I1-I11 invariants, frontmatter requirements) and a marketplace.json description + $schema

No behavioral changes. claude plugin validate passes clean.

Author: tobinsouth

.claude-plugin/marketplace.json

@@ -1,5 +1,7 @@
 {
+  "$schema": "https://anthropic.com/claude-code/marketplace.schema.json",
   "name": "claude-for-legal",
+  "description": "Reference agents, skills, and data connectors for the legal workflows we see most — in-house commercial, privacy, product, corporate, employment, litigation, regulatory, AI governance, IP, and the learning side of the practice (law school clinics and students).",
   "owner": {
     "name": "Anthropic"
   },
@@ -95,7 +97,7 @@
     {
       "name": "ip-legal",
       "source": "./ip-legal",
-      "description": "Runs first-pass trademark clearance and freedom-to-operate triage, drafts and triages cease-and-desist letters and DMCA takedowns (send and respond), checks open source compliance, reviews IP clauses, and tracks registrations and renewal deadlines.",
+      "description": "Runs first-pass trademark clearance and freedom-to-operate triage, screens invention disclosures for initial patentability, drafts and triages cease-and-desist letters and DMCA takedowns (send and respond), checks open source compliance, reviews IP clauses, and tracks registrations and renewal deadlines.",
       "author": {
         "name": "Anthropic"
       }

CLAUDE.md

@@ -0,0 +1,130 @@
+# CLAUDE.md
+
+Guidance for working on this repo. `claude-for-legal` is a Claude Code plugin
+marketplace — twelve first-party legal plugins, one vendor plugin, and five
+managed-agent cookbooks. Most work here is editing prompt content (skills,
+agents, hooks), plugin metadata, or cookbook config — not application code.
+
+## Layout
+
+```
+.claude-plugin/marketplace.json   # the marketplace manifest — one entry per plugin
+<plugin>/                         # 12 first-party plugins (commercial-legal, privacy-legal, ...)
+  .claude-plugin/plugin.json      # plugin manifest (name, version, description, author)
+  .mcp.json                       # MCP servers the plugin connects to
+  CLAUDE.md                       # practice-profile TEMPLATE (see "Plugin CLAUDE.md" below)
+  README.md                       # per-plugin docs
+  skills/<name>/SKILL.md          # one skill per directory
+  agents/<name>.md                # subagent definitions
+  hooks/hooks.json                # hook config (most plugins ship an empty stub)
+  .gitignore
+external_plugins/<vendor>/        # vendor-maintained plugins (CoCounsel)
+managed-agent-cookbooks/<name>/   # CMA agent.yaml + subagents/ + steering-examples.json
+scripts/                          # validate.py, lint-tool-scope.py, orchestrate.py,
+                                  # deploy-managed-agent.sh, test-cookbooks.sh
+references/                       # shared templates (company-profile, dashboard)
+```
+
+## Validation — run before opening a PR
+
+This repo follows the same conventions `anthropics/claude-plugins-official`
+enforces in CI. Run the equivalent checks locally:
+
+```bash
+# 1. Marketplace + per-plugin schema validation (source of truth)
+claude plugin validate .claude-plugin/marketplace.json
+for d in */; do [ -f "$d/.claude-plugin/plugin.json" ] && claude plugin validate "$d"; done
+claude plugin validate external_plugins/cocounsel-legal
+
+# 2. Cookbook tool-scope lint (orchestrators must not over-grant tools)
+python3 scripts/lint-tool-scope.py
+
+# 3. JSON/YAML sanity
+python3 -c "import json,glob; [json.load(open(f)) for f in glob.glob('**/*.json', recursive=True)]"
+```
+
+### Marketplace invariants (I1–I11)
+
+`claude-plugins-official` layers these on top of the schema check. They apply
+here too — the ones most likely to trip a contributor:
+
+- **I1** — `plugins[]` should be alpha-sorted by name (case-insensitive).
+  *Currently a known warning: the array is in a curated display order. If you
+  add a plugin, ask before re-sorting the whole array.*
+- **I2** — no duplicate plugin names.
+- **I3** — `description` 10–2000 chars, no leading/trailing whitespace.
+- **I8** — every vendored `source` (`"./<dir>"`) must point at a directory that
+  contains `.claude-plugin/plugin.json`.
+- **I9** — `source` paths/URLs must contain no shell metacharacters or `..`.
+- **I10** — no hidden Unicode (zero-width chars, bidi controls) in
+  `name`/`description`.
+- **I11** — `name` must match `^[a-z0-9][a-z0-9-]{1,63}$`.
+
+### Frontmatter requirements
+
+Every `agents/*.md` needs `name` and `description`. Every
+`skills/<name>/SKILL.md` needs `description`. Every `commands/*.md` needs
+`description`. Multi-line descriptions use `>` block scalars and that's fine —
+`claude plugin validate` parses them correctly.
+
+## Conventions
+
+### Keep `marketplace.json` in sync with `plugin.json`
+
+For first-party plugins, `marketplace.json`'s `name`, `description`, and
+`author` should match the plugin's own `.claude-plugin/plugin.json` field for
+field. If you change a plugin's description in one place, change it in the
+other.
+
+### Skill names in prose must be canonical
+
+When a `SKILL.md` (especially `customize` or `cold-start-interview`) tells the
+user "run `/foo`," `foo` must be the actual `skills/<foo>/` directory name.
+Short forms like `/triage` for `/use-case-triage` look right in prose but are
+dead commands — the user types them and nothing happens. Refs to Claude Code
+built-ins (`/mcp`, `/plugin`) and to other plugins (`/<other-plugin>:<skill>`)
+are fine.
+
+### Plugin CLAUDE.md is a template, not project context
+
+Each `<plugin>/CLAUDE.md` is a practice-profile template that the
+`cold-start-interview` skill copies to `~/.claude/plugins/config/claude-for-legal/<plugin>/CLAUDE.md`
+on the user's machine. It is *not* loaded as project context when the plugin is
+installed — `claude plugin validate` warns about this and the warning is
+expected. Don't "fix" it by moving the content into a skill.
+
+### `external_plugins/` is vendor-maintained
+
+Plugins under `external_plugins/` are built and maintained by the vendor
+(README.md has the policy). Don't change vendor-authored content without
+checking with them first; whitespace normalization and formatting are usually
+fine since the vendor lands changes via PR rather than mirroring a fork.
+
+### Formatting
+
+- 2-space indent in all JSON and `.mcp.json` files.
+- Final newline at end of every text file.
+- No trailing whitespace.
+- Markdown tables: pipe-aligned columns are nice but not required; just keep
+  the column count consistent.
+
+## Cookbooks
+
+Each `managed-agent-cookbooks/<name>/` has `agent.yaml` (the orchestrator),
+`subagents/*.yaml` (the leaves), `steering-examples.json`, and `README.md`. Two
+rules that `scripts/lint-tool-scope.py` enforces:
+
+1. The orchestrator gets local-only tools (`read`, `grep`, `glob`,
+   `agent_toolset`); MCP and write tools belong to specific subagent leaves.
+2. The README's security table and the `agent.yaml` comments must match what
+   the YAML actually grants. Don't claim a tool a subagent doesn't have.
+
+## Things to leave alone
+
+- Per-plugin `.gitignore` files differ slightly across plugins. Probably
+  intentional; ask before unifying.
+- `hooks/hooks.json` is missing in two plugins. Hooks are optional; the missing
+  files are not a bug.
+- `references/` lives only at repo root and is not shipped inside any plugin
+  directory. Several plugin `CLAUDE.md` templates reference it as if it were —
+  that's a known gap, not a thing to silently move.

README.md

@@ -150,7 +150,7 @@ managed-agent-cookbooks/  # Claude Managed Agent cookbooks — one dir per sched
   launch-radar/
   reg-monitor/
   renewal-watcher/
-scripts/                  # deploy-managed-agent.sh · validate.py · orchestrate.py · lint-tool-scope.py
+scripts/                  # deploy-managed-agent.sh · validate.py · orchestrate.py · lint-tool-scope.py · test-cookbooks.sh
 .claude-plugin/
   marketplace.json        # plugin registry
 ```
@@ -361,6 +361,7 @@ The full map across all plugins. The cold-start interview is the first thing to
 | Command | Skill | What it does |
 |---|---|---|
 | `/ai-governance-legal:cold-start-interview` | cold-start-interview | Cold-start — learns your AI governance practice |
+| `/ai-governance-legal:ai-inventory` | ai-inventory | EU AI Act per-system inventory — track each system's role and risk tier |
 | `/ai-governance-legal:use-case-triage` | use-case-triage | Classify AI use case — approved, conditional, or no |
 | `/ai-governance-legal:aia-generation` | aia-generation | Run an AI impact assessment in house format |
 | `/ai-governance-legal:vendor-ai-review` | vendor-ai-review | Review vendor AI terms against governance positions |
@@ -465,6 +466,7 @@ The full map across all plugins. The cold-start interview is the first thing to
 | `/ip-legal:cold-start-interview` | cold-start-interview | Cold-start — learn your IP practice and posture |
 | `/ip-legal:clearance` | clearance | Trademark clearance first pass — knockout + similar marks |
 | `/ip-legal:fto-triage` | fto-triage | Freedom-to-operate triage, not an FTO opinion |
+| `/ip-legal:invention-intake` | invention-intake | Invention disclosure first-pass screen — novelty, obviousness, §101, bar dates |
 | `/ip-legal:cease-desist` | cease-desist | Draft a C&D or triage one you received |
 | `/ip-legal:takedown` | takedown | DMCA notice, response triage, or §512(g) counter-notice |
 | `/ip-legal:infringement-triage` | infringement-triage | Infringement triage across all four IP rights |

ai-governance-legal/.claude-plugin/plugin.json

@@ -5,4 +5,4 @@
   "author": {
     "name": "Anthropic"
   }
-}
+}

ai-governance-legal/README.md

@@ -32,6 +32,7 @@ positions and house style.
 | Command | Does |
 |---|---|
 | `/ai-governance-legal:cold-start-interview` | Cold-start interview — writes your practice profile |
+| `/ai-governance-legal:ai-inventory [list \| add \| edit \| classify \| show]` | Manage the EU AI Act per-system inventory — track each system's role and risk tier |
 | `/ai-governance-legal:use-case-triage [use case]` | Classify a use case against your registry (approved / conditional / never) |
 | `/ai-governance-legal:aia-generation [use case]` | Run an AI impact assessment (AIA) in your house style |
 | `/ai-governance-legal:vendor-ai-review [vendor/file]` | Review a vendor AI agreement against your positions |
@@ -45,6 +46,7 @@ positions and house style.
 | Skill | Purpose |
 |---|---|
 | **cold-start-interview** | Writes `~/.claude/plugins/config/claude-for-legal/ai-governance-legal/CLAUDE.md` from interview + seed docs |
+| **ai-inventory** | EU AI Act per-system inventory — role (provider, deployer, importer, distributor, authorized rep, product manufacturer) and risk tier per system |
 | **use-case-triage** | Classifies use cases against the registry; flags missing assessments |
 | **aia-generation** | AI impact assessment (AIA) in house format |
 | **vendor-ai-review** | AI-specific vendor contract review against governance positions |

ai-governance-legal/skills/cold-start-interview/SKILL.md

@@ -269,7 +269,7 @@ that's where to go deep first.
 ### Part 2: Regulatory footprint (2-3 min)
 
 > Which regulations are actually on your radar? I don't want to assume — tell me
-> what's real for you. (This feeds /gap-check and /policy-monitor — the gap analysis diffs new regulations against your stated scope, and policy-monitor only watches regimes you've marked in scope.)
+> what's real for you. (This feeds /reg-gap-analysis and /policy-monitor — the gap analysis diffs new regulations against your stated scope, and policy-monitor only watches regimes you've marked in scope.)
 
 **Do not assume any regulation applies. Ask the user which regimes they think apply, then research the AI-specific regulations currently in effect or pending in the jurisdictions where the company operates, deploys AI, or has affected parties. This landscape changes quickly — verify currency.**
 
@@ -336,7 +336,7 @@ walk through each one.
 
 **The red lines question:**
 > "What's the use case that's an automatic no — the thing someone could propose
-> and you'd stop them immediately without needing to think about it?" (This feeds /triage — the skill checks proposed AI use cases against these red lines before doing anything else, and flags anything on the list as automatic stop.)
+> and you'd stop them immediately without needing to think about it?" (This feeds /use-case-triage — the skill checks proposed AI use cases against these red lines before doing anything else, and flags anything on the list as automatic stop.)
 
 Common categories to probe if they're slow: biometric data, emotion detection,
 political/religious inference, fully automated adverse decisions affecting employment
@@ -360,7 +360,7 @@ or credit, uses involving children.
 
 **Escalation:**
 
-> "When a review finds something that needs someone more senior to sign off — a vendor AI agreement with training-on-data or liability issues, an AI use case that doesn't fit your registry, a regulatory gap that needs a decision, or a call above your authority — who does that go to? Give me a name or a role (the GC, the Chief Privacy Officer, your boss), or say 'I decide myself.' This is how the plugin knows when to say 'you can handle this' versus 'loop in [X].' (This feeds every skill's routing logic — /triage, /review-vendor-ai, and /gap-check all check the escalation matrix before telling you to hand something up.)"
+> "When a review finds something that needs someone more senior to sign off — a vendor AI agreement with training-on-data or liability issues, an AI use case that doesn't fit your registry, a regulatory gap that needs a decision, or a call above your authority — who does that go to? Give me a name or a role (the GC, the Chief Privacy Officer, your boss), or say 'I decide myself.' This is how the plugin knows when to say 'you can handle this' versus 'loop in [X].' (This feeds every skill's routing logic — /use-case-triage, /vendor-ai-review, and /reg-gap-analysis all check the escalation matrix before telling you to hand something up.)"
 
 Also ask:
 - Has anything been escalated to the board or C-suite over AI in the last year?
@@ -374,7 +374,7 @@ Also ask:
 ### Part 5: Seed documents (3-4 min)
 
 > "I want to see what you actually have. Tell me which of these exist, and share
-> what you can. (The AI policy feeds /policy-monitor drift detection; the prior impact assessment becomes the /aia-generation template; the vendor agreements become the starting playbook for /review-vendor-ai.)"
+> what you can. (The AI policy feeds /policy-monitor drift detection; the prior impact assessment becomes the /aia-generation template; the vendor agreements become the starting playbook for /vendor-ai-review.)"
 >
 > 1. **AI or acceptable use policy.** Your internal or public-facing policy on how
 >    AI can and can't be used. This tells me your committed positions.

ai-governance-legal/skills/customize/SKILL.md

@@ -70,9 +70,9 @@ cold-start interview and without hand-editing YAML.
      conditional rather than approved, surface more AIA follow-ups, and
      recommend more conservative vendor AI redlines."
    - *Adding an escalation contact:* "Every skill that routes escalations
-     (`/triage`, `/review-vendor-ai`, `/gap-check`) will now include this
+     (`/use-case-triage`, `/vendor-ai-review`, `/reg-gap-analysis`) will now include this
      contact on the relevant risk bands."
-   - *New use case registry entry:* "`/triage` will match against this entry
+   - *New use case registry entry:* "`/use-case-triage` will match against this entry
      on its next run. Existing AIAs aren't rewritten — re-run them if you want
      the new posture reflected."
 
@@ -92,7 +92,7 @@ cold-start interview and without hand-editing YAML.
 
 - **Never delete a section.** If the user wants to "remove" something, set it
   to `[Not configured]` and explain what that means for the plugin's behavior.
-  ("Removing your escalation chain means `/triage` will flag escalation-worthy
+  ("Removing your escalation chain means `/use-case-triage` will flag escalation-worthy
   items but won't route them to a specific person.")
 - **Flag internal inconsistency.** If the change would make the profile
   inconsistent (e.g., risk posture aggressive + escalation "everything goes to

commercial-legal/.claude-plugin/plugin.json

@@ -5,4 +5,4 @@
   "author": {
     "name": "Anthropic"
   }
-}
+}

commercial-legal/skills/cold-start-interview/SKILL.md

@@ -173,7 +173,7 @@ Two quick questions before we get into commercial-contracts specifics. These sha
 
 #### Who's using this?
 
-> Who'll be using this plugin day to day? (This feeds the work-product header on every /review, /amendment-history, and /renewals-due output — lawyer gets "PRIVILEGED & CONFIDENTIAL — ATTORNEY WORK PRODUCT"; non-lawyer gets "RESEARCH NOTES — NOT LEGAL ADVICE" plus research-framed outputs.)
+> Who'll be using this plugin day to day? (This feeds the work-product header on every /review, /amendment-history, and /renewal-tracker output — lawyer gets "PRIVILEGED & CONFIDENTIAL — ATTORNEY WORK PRODUCT"; non-lawyer gets "RESEARCH NOTES — NOT LEGAL ADVICE" plus research-framed outputs.)
 >
 > 1. **Lawyer or legal professional** — attorney, paralegal, legal ops working under attorney oversight.
 > 2. **Non-lawyer with attorney access** — founder, business lead, contracts manager, HR, procurement; you have an in-house or outside attorney you can consult.
@@ -347,7 +347,7 @@ If they don't have one: proceed with the questions below.
 
 **Approval levels**
 
-> When a review finds something that needs someone more senior to sign off — a term that's above playbook (a higher LoL cap, an indemnity structure outside your fallbacks), a risk that needs a second opinion, or a decision that's above your authority — who does that go to? Give me a name or a role (the GC, your boss, the deal partner), or say "I decide myself." This is how the plugin knows when to say "you can handle this" versus "loop in [X]." (This feeds /escalate — the skill drafts the escalation ask using this matrix, and /review uses it to decide whether a flagged term lands in your lane or somebody else's.)
+> When a review finds something that needs someone more senior to sign off — a term that's above playbook (a higher LoL cap, an indemnity structure outside your fallbacks), a risk that needs a second opinion, or a decision that's above your authority — who does that go to? Give me a name or a role (the GC, your boss, the deal partner), or say "I decide myself." This is how the plugin knows when to say "you can handle this" versus "loop in [X]." (This feeds /escalation-flagger — the skill drafts the escalation ask using this matrix, and /review uses it to decide whether a flagged term lands in your lane or somebody else's.)
 
 **Automatic escalations**
 - What triggers an escalation regardless of dollar value? (Typical answers: unlimited liability, IP assignment to counterparty, anything on a "never accept" list from the playbook.)

commercial-legal/skills/customize/SKILL.md

@@ -67,7 +67,7 @@ cold-start interview and without hand-editing YAML.
      anything above 6 months as a deviation; existing deal-debrief entries
      stay as logged."
    - *New escalation approver:* "Any redline exceeding your own authority
-     will now route to this approver — `/escalate` will include them by
+     will now route to this approver — `/escalation-flagger` will include them by
      default for the matching risk band."
    - *Risk posture middle → aggressive:* "I'll accept more vendor-friendly
      positions without flagging them and shift the `[review]` bar higher."