Skip to content

docs(skill): whole-trip-planning mna skill + Claude Code install - #3

Merged
akoso merged 1 commit into
mainfrom
docs/mna-skill-playbook
Jun 13, 2026
Merged

akoso merged 1 commit into
mainfrom
docs/mna-skill-playbook

Conversation

@akoso

@akoso akoso commented Jun 13, 2026

Copy link
Copy Markdown
Collaborator

Rewrites the mna Claude Code skill from the outdated read-only "Phase 0" stub into a reusable, whole-trip-planning playbook, and adds a one-command install.

What changed

  • skills/mna/SKILL.md — data model, a set of collaboration principles (app as source of truth + local PLAN.md mirror, recommend-don't-enumerate, verify real availability before recommending, cost with looked-up numbers, honor the user's manual edits, verify-after-write), and a step-by-step planning playbook.
  • skills/mna/references/cli-and-schemas.md — command map, option/event JSON body shapes + enums, the nested-coordinates location requirement, date/cancellation fields, verify-after-write.
  • skills/mna/references/research-and-costing.md — hotel-connector research with exact-date availability verification, and own-car fuel costing.
  • README.md — one-command Claude Code skill install; reflect that writes are live; link the skill; drop the stale roadmap note.

Install (also in the README)

git clone --depth 1 https://github.com/akoso/mna-cli /tmp/mna-cli && mkdir -p ~/.claude/skills && cp -r /tmp/mna-cli/skills/mna ~/.claude/skills/

Notes

  • Generic & public — no trip-identifying details; lessons stated as reusable conclusions.
  • Connector-agnostic — works with any hotel connector or plain web search; Booking.com is named only as a recommendation.
  • Branched off main, independent of the date/free-cancellation feature branch.

🤖 Generated with Claude Code

The old SKILL.md described a read-only Phase-0 CLI; the CLI now does full
trip mutation. Replace it with a general, reusable planning method:

- SKILL.md: data model, collaboration principles (app as source of truth,
  recommend-don't-enumerate, verify real availability, cost with looked-up
  numbers, honor user edits, verify-after-write) and a step-by-step playbook.
- references/cli-and-schemas.md: command map, option/event JSON body shapes,
  enums, the nested-coordinates location requirement, dates, verify-after-write.
- references/research-and-costing.md: hotel-connector research + exact-date
  availability verification, and own-car fuel costing.
- README: one-command Claude Code skill install; writes are live; link the
  skill; drop the stale roadmap note.

Generic and connector-agnostic (any hotel connector or web search).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@akoso
akoso merged commit ca2881c into main Jun 13, 2026
1 check passed
akoso added a commit that referenced this pull request Jul 29, 2026
Prepares the CLI for its first npm release under the @mantacode scope.

Rename: the package becomes `@mantacode/mna-cli` (was `@mynextadventure/cli`).
The scope holds multiple products, so the package name carries the product
prefix; the installed binary stays `mna`. Homebrew references now point at
`mantacode/homebrew-tap` / `brew install mantacode/tap/mna`.

Packaging hardening: added `publishConfig.access: public` alongside the
workflow's existing `--access public` (scoped packages are private by
default, which bites hardest on the first publish), plus repository/homepage/
bugs/keywords metadata for the npm page.

RELEASING.md now states the first-release version explicitly (0.1.0, not
1.0.0 — 1.0.0 is reserved as the signal that the /v1/ contract is frozen)
and warns that NPM_TOKEN must exist before the tag is pushed, since a failed
publish burns the version number.

DEVELOPMENT.md documented the codegen source priority backwards: it claimed
the live URL is tried first with the committed snapshot as fallback, but
scripts/codegen.ts tries the snapshot first. It also still called releasing
"out of scope until Phase 8" and pointed at a design doc in another repo.

README: fixed a `git clone` URL for an org that doesn't exist, and the
paste-token host (app.mynextadventure.cloud).

skills/mna/ is left alone — PR #3 already truth-passed it.

Version deliberately left at 0.0.1; RELEASING.md owns the bump.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AsdP2hvt4XQhYxkdNREXUq
akoso added a commit that referenced this pull request Jul 29, 2026
…#4)

Prepares the CLI for its first npm release under the @mantacode scope.

Rename: the package becomes `@mantacode/mna-cli` (was `@mynextadventure/cli`).
The scope holds multiple products, so the package name carries the product
prefix; the installed binary stays `mna`. Homebrew references now point at
`mantacode/homebrew-tap` / `brew install mantacode/tap/mna`.

Packaging hardening: added `publishConfig.access: public` alongside the
workflow's existing `--access public` (scoped packages are private by
default, which bites hardest on the first publish), plus repository/homepage/
bugs/keywords metadata for the npm page.

RELEASING.md now states the first-release version explicitly (0.1.0, not
1.0.0 — 1.0.0 is reserved as the signal that the /v1/ contract is frozen)
and warns that NPM_TOKEN must exist before the tag is pushed, since a failed
publish burns the version number.

DEVELOPMENT.md documented the codegen source priority backwards: it claimed
the live URL is tried first with the committed snapshot as fallback, but
scripts/codegen.ts tries the snapshot first. It also still called releasing
"out of scope until Phase 8" and pointed at a design doc in another repo.

README: fixed a `git clone` URL for an org that doesn't exist, and the
paste-token host (app.mynextadventure.cloud).

skills/mna/ is left alone — PR #3 already truth-passed it.

Version deliberately left at 0.0.1; RELEASING.md owns the bump.


Claude-Session: https://claude.ai/code/session_01AsdP2hvt4XQhYxkdNREXUq

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant