Skip to content
Back to skills

065 Lower Vocabulary Barrier

ASecurity

**Goal:** Give a junior an on-demand "explain it to me like I'm new here" command — `/jig:explain <term>` defines a single term in plain language; `/jig:explain <spec-or-adr-path>` produces a strong-handholding walkthrough of a dense artifact, auto-pulling the ADRs/specs it links so the reader doesn't chase references. Output is ephemeral (chat-only).

  • 7 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 4, 2026
ai-agentspythongo

Security analysis

A100/100

Pro scans all 21 files and shows the line behind each finding

Scanned September 4, 2026

npx -y skills add ramboz/jig --skill 065-lower-vocabulary-barrier --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of 065 Lower Vocabulary Barrier?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for 065 Lower Vocabulary Barrier
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/ramboz-065-lower-vocabulary-barrier/badge)](https://www.skillsdirectory.com/skills/ramboz-065-lower-vocabulary-barrier)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
status: DONE
dependencies: [065-01]
last_verified: 2026-06-07
arch_review: true  # adds a new public skill (a new external surface in the
#                    plugin manifest + CLAUDE.md skills table).
---

## Slice 065-03 — `/jig:explain` skill (term + artifact modes)

**Goal:** Give a junior an on-demand "explain it to me like I'm new here" command
— `/jig:explain <term>` defines a single term in plain language; `/jig:explain
<spec-or-adr-path>` produces a strong-handholding walkthrough of a dense
artifact, auto-pulling the ADRs/specs it links so the reader doesn't chase
references. Output is ephemeral (chat-only).

**DoR:**
- ✅ 065-01 landed — the merged lexicon is the source for term-mode definitions
  and for the "words you'll need" inventory in artifact mode.
- ✅ The judgment-skill pattern is proven (`/jig:clarify`, `/jig:pr-review`,
  `/jig:arch-review` — SKILL.md only, no `.py` helper).
- ✅ Plugin manifest + `validate_manifests.py` + the CLAUDE.md skills table are
  the registration surfaces a new skill must satisfy.

**Acceptance Criteria:**

1. **The skill is registered and discoverable.** `skills/explain/SKILL.md`
   exists, is listed in the plugin manifest (passing `validate_manifests.py`),
   and has a row in the CLAUDE.md "Skills in this repo" table. Its description
   declares both invocation styles (auto + explicit) and the two modes.
2. **Term mode.** The SKILL.md specifies that `/jig:explain <term>` returns the
   term's plain-language definition from the merged lexicon (065-01), with its
   example + see-also when present, and **flags** a term absent from the lexicon
   (rather than inventing a definition).
3. **Artifact mode — strong handholding.** The SKILL.md specifies that
   `/jig:explain <path-to-spec-or-adr>` produces a junior-grade walkthrough with a
   fixed shape: **In one sentence** / **Why it exists** / **Words you'll need
   first** (every jig term in the artifact, defined inline; gaps flagged) /
   **Walkthrough** (section by section, plain language) / **The decisions & why**
   (for ADRs: alternatives + trade-off) / **If you had to work on this**. It
   **auto-pulls linked ADRs/specs** so references are resolved for the reader.
4. **Ephemeral.** The skill writes nothing to disk — output is chat-only. The
   SKILL.md states this explicitly (no `--save`, no appended section), keeping the
   hot path clean (clarify Q3 / 055/057).
5. **Judgment-only, no helper.** No `.py` is added for this skill; section
   surgery / lookups happen inline via Read + the 065-01 loader.
6. **Defers to a richer installed skill.** Per the jig baseline pattern, the
   description defers to any other installed skill whose description identifies it
   as handling plain-language explanation / onboarding / artifact walkthroughs —
   prefer it over this slim baseline (and does not defer to the generic built-in).

_Testability note (accepted gap): ACs 1, 4, 5, 6 are structural and unit-tested
(manifest registration, ephemeral contract, no-helper, deferral language). ACs 2
and 3 govern output **quality**, which is judgment exercised by the skill prompt,
not a unit test — the same accepted shape as every judgment-only jig skill
(`/jig:clarify`, `/jig:pr-review`). Recorded in the spec's coverage summary._

**DoD:**
- [x] All ACs pass; full test suite green (manifest validation included).
- [x] Coverage: `validate_manifests.py` passes with the new skill; a test asserts
      the SKILL.md declares term + artifact modes, the ephemeral contract, and the
      deferral language; the CLAUDE.md skills-table row is present.
- [x] Reviewed by `reviewer` subagent. Reviewer prompt built by `review.py`.
- [x] Implementation review passed.
- [x] Craft (pr-review) pass run; blockers addressed.
- [x] Arch (arch-review) pass run (slice declares `arch_review: true`); blockers addressed.
- [x] Deviation log produced under this slice heading.
- [x] `docs/refinement-todo.md` updated if any decisions were deferred. (No decisions
      deferred; one non-blocking follow-up noted in the deviation log.)

**Anti-horizontal-phasing check:** After this slice, a junior can run
`/jig:explain docs/specs/062-refactor-workflow/spec.md` and get a plain-language
walkthrough with linked ADRs pulled in — a complete, usable capability, not a
stub.

### Close-out (post-DONE)

- [x] `docs/specs/README.md` regenerated; Notes column records: `/jig:explain`
      = term + artifact modes, ephemeral, defers to richer installed skill.
- [x] `CLAUDE.md` hygiene per spec 025-01: add the `/jig:explain` row to the
      Skills table; leave the Active-specs entry until the closing slice.

### Deviation log (after reconciliation)

The original spec is preserved above. Implementation notes:

- **Deliverables.** `skills/explain/SKILL.md` (the judgment-skill prompt, no
  `.py` helper) + `skills/explain/test_explain_skill_surface.py` (27 structural
  surface tests). Registered across the three tier tables —
  `skills/scaffold-init/scaffold.py` `_TIER_SKILLS["tier-1"]` (source of truth),
  `scripts/install_contract.py` `EXPECTED_SKILLS`, `scripts/scaffold_contract.py`
  `_TIER_SKILLS["tier-1"]` — plus the `/jig:explain` row in the root `CLAUDE.md`
  Skills table.

- **AC1 phrasing vs. jig's actual discovery mechanism (no behavior change).**
  AC1 says the skill must be "listed in the plugin manifest (passing
  `validate_manifests.py`)". jig does **not** enumerate skills in a manifest —
  skills are directory-auto-discovered, and the install contract carries them via
  `scaffold._TIER_SKILLS` (source of truth) + the two restated mirror tables
  (`install_contract.EXPECTED_SKILLS`, `scaffold_contract._TIER_SKILLS`, pinned
  equal by the consistency tests). `validate_manifests.py` validates the three
  JSON manifests (plugin/marketplace/hooks — still passing) but does not list
  skills. The implementation satisfies AC1's **intent** (registered +
  discoverable, install contract carries it) through those real registration
  surfaces; the test (`TierRegistrationTests`) asserts the actual surfaces, not a
  literal manifest entry. Flagged independently by all three reviewers.

- **Term-mode recipe made layout-agnostic (reconciliation fix).** All three
  review passes raised the same non-blocking nit: the inline `python3 -c` lexicon
  recipe in term mode hardcoded `sys.path.insert(0, 'skills/_common')` with the
  scaffolded `.claude/skills/_common` path only in a prose aside — not
  copy-paste-safe in a scaffolded project, and not auto-corrected by the 046-01
  rewrite policer (which targets `${CLAUDE_PLUGIN_ROOT}/skills/...` helper paths,
  not relative `python3 -c` snippets). Fixed during reconciliation: the recipe
  now probes both `skills/_common` and `.claude/skills/_common` and inserts the
  one that exists. Surface tests re-run green (27/27).

- **Open, non-blocking follow-up (not deferred — no decision needed).** The
  craft pass noted `TierRegistrationTests` mutate `sys.path` via `insert(0, …)`
  without cleanup. Harmless for the current suite (the inserted dirs host
  uniquely-named jig-internal modules), left as-is to match the existing
  surface-test idiom; a future sweep could wrap them in `try/finally`. No
  refinement-todo entry — it's a test-hygiene nicety, not an undefined decision.

- **Pinned-tier-set guards + tier-inventory docs updated (by design).** Adding
  `explain` to tier-1 tripped jig's deliberate "tier addition → update the
  pinned set / docs" guards, all updated to include it:
  `test_scaffold.TierSkillSetTests.test_tier_1_is_pinned` (`EXPECTED_TIER_1`) +
  `test_migrate.TierUpgradeTests` (`TIER1`) pin the scaffolded/migrated tier-1
  set; `test_product_vision_names_every_tier_skill` forced the
  `docs/product-vision.md` Tier-1 inventory entry (#17) + its "7 Tier 0 + 10
  Tier 1" headline count; `test_vision_elicitation_worked_example_tier_line_in_sync`
  forced the same count + skill-list updates in
  `skills/vision-elicitation/worked-example-jig.md`. These are exactly the
  deliberate-update guards firing, distinct from the `TierRegistrationTests`
  that pin the three source tables equal.

- **All three review passes (compliance / craft / arch) returned `pass`**;
  verdict artifacts recorded under `reviews/slice-03-{compliance,craft,arch}.md`;
  reconciliation pass `pass` at `reviews/slice-03-reconciliation.md`.

Files in this skill

  • reviews/slice-01-arch.md1.7 KB
  • reviews/slice-01-compliance.md1.2 KB
  • reviews/slice-01-craft.md1.5 KB
  • reviews/slice-01-reconciliation.md1.1 KB
  • reviews/slice-02-compliance.md1.4 KB
  • reviews/slice-02-craft.md1.6 KB
  • reviews/slice-02-reconciliation.md924 B
  • reviews/slice-03-arch.md1.6 KB
  • reviews/slice-03-compliance.md1.5 KB
  • reviews/slice-03-craft.md1.5 KB
  • reviews/slice-03-reconciliation.md1.1 KB
  • reviews/slice-04-arch.md1.5 KB
  • reviews/slice-04-compliance.md1.1 KB
  • reviews/slice-04-craft.md1.2 KB
  • reviews/slice-04-reconciliation.md1.2 KB
  • reviews/slice-05-arch.md1.8 KB
  • reviews/slice-05-compliance.md1.3 KB
  • reviews/slice-05-craft.md1.2 KB
  • reviews/slice-05-reconciliation.md1.1 KB
  • slice-01-lexicon-foundation.md8.4 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.