skills/dev-needs-panel/SKILL.md
--- name: dev-needs-panel description: At the end of an autonomous run, generate a self-contained HTML "what I need from you" panel the user opens when back at the keyboard. Shows per-project / per-worktree what the agent is blocked on (env vars, accounts, config) with fillable form fields and a Save button that writes a local answers file the agent reads next session (fallback - the user just says "updated"). Generalises the completion-report + telegram-ping pattern into an actionable inbound c
npx skillsauth add RonanCodes/ronan-skills skills/dev-needs-panelInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
The outbound sibling of completion-report. Completion-report answers "what did I do". This answers "what do I need from you before I can keep going". One self-contained HTML file per project / worktree, opened when the user is back at the keyboard, with fillable fields and a Save that the agent reads back next session.
This exists because the expensive blocker in an autonomous run is a missing secret or an account the agent can't create. Telling the user about it in a chat transcript they have already scrolled past loses it. A single browsable panel that survives the session, names the exact env vars / accounts / config, and lets the user paste values into it turns a dead run into a queued one.
/ro:night-shift, /ro:ralph, or /ro:planner-worker run that stalled
on something only the human can provide (a Stripe key, a confirmed account, a config value).--url file://...).Skip it when there are no outstanding needs. An empty panel is noise.
scripts/generate.sh renders that spec into one self-contained .html file. No CDN, no
external assets - all CSS and JS are inlined.<panel>.answers.json) next to the panel. The agent
reads that file next session to pick up the values.file:// open without the File
System Access API), Save downloads the JSON and shows the exact path to drop it. Fallback to
that: the user just tells the agent "updated" and the agent reads whatever the user pasted in.Pipe the spec on stdin, or pass --spec <file>:
bash ~/Dev/ronan-skills/skills/dev-needs-panel/scripts/generate.sh \
--out .dev-needs/needs.html <<'SPEC'
{
"project": "makely",
"worktree": "/tmp/ns-wt/makely-2",
"branch": "feat/checkout",
"links": [
{"label": "Repo", "url": "https://github.com/RonanCodes/makely"},
{"label": "PR #12", "url": "https://github.com/RonanCodes/makely/pull/12"}
],
"needs": [
{
"key": "STRIPE_SECRET_KEY",
"label": "Stripe secret key",
"kind": "env",
"why": "Checkout flow can't be wired without it.",
"where": "~/.claude/.env",
"secret": true,
"required": true
},
{
"key": "stripe_account",
"label": "Stripe account confirmed",
"kind": "account",
"why": "Need a test account to create products.",
"type": "checkbox"
}
]
}
SPEC
The script prints the absolute path of the written HTML on its last line. With no --out it
writes to .dev-needs/<timestamp>-<project>.html in the current repo.
Open it with --open (calls open on macOS), or just hand the user the path / deep-link it
from the end-of-run ping.
Top-level object:
| Field | Required | Meaning |
|------------|----------|---------|
| project | yes | Project name, shown in the header and used for the filename slug. |
| worktree | no | Worktree path, so it's unambiguous which checkout this is for. |
| branch | no | Branch name, shown in the header. |
| links | no | Array of {label, url} - repo, PR, dashboard. Rendered as link buttons. |
| needs | yes | Array of need objects (below). An empty array renders an empty-state panel. |
Each need object:
| Field | Required | Default | Meaning |
|------------|----------|---------|---------|
| key | yes | - | Stable identifier the agent reads back from the answers file. |
| label | no | key | Human label on the field. |
| kind | no | other | env | account | config | other. Shown as a colour tag. |
| why | no | - | One line on why the agent is blocked on it. |
| where | no | - | Where the value should live (e.g. ~/.claude/.env). |
| secret | no | false | Render as a password field. |
| required | no | false | Mark the field required (badge + HTML required). |
| type | no | text | text | textarea | checkbox. |
Save writes (or downloads) a JSON sibling of the HTML, named <panel-basename>.answers.json:
{
"meta": {
"project": "makely",
"worktree": "/tmp/ns-wt/makely-2",
"branch": "feat/checkout",
"answersFile": "20260628-145541-makely.answers.json",
"answersPath": "/abs/path/.dev-needs/20260628-145541-makely.answers.json",
"generatedAt": "20260628-145541"
},
"answers": {
"STRIPE_SECRET_KEY": "sk_test_...",
"stripe_account": true
},
"savedAt": "2026-06-28T13:00:00.000Z"
}
answers is keyed by each need's key. Checkboxes are booleans; everything else is a string.
Next session, before re-running the blocked work:
.dev-needs/*.answers.json in the repo, newest first.answers object. Apply each value where its need's where said it should go
(e.g. append KEY=value to ~/.claude/.env, never commit secrets).~/.claude/.env) and proceed.The .dev-needs/ directory is per-run, machine-local, and may contain secrets the user pasted.
It must be gitignored. Add this to the target repo's .gitignore if missing:
.dev-needs/
Never commit a panel or an answers file.
--url file://<panel-path> so tapping the phone
notification opens the panel.skills/completion-report/SKILL.md and skills/telegram/SKILL.md for the shared style;
do not edit them.testing
--- name: linear-pipeline description: The Fable orchestrator for a single dispatched Linear ticket. Holds almost no context itself; it receives `--issue <ID> --detached`, decides the stage sequence, and fans out a sub-agent per stage, passing forward only each stage's artifact (never re-derived, never inlined into its own context). Step zero, before any planning or stage routing, is a boundary triage against `canon/security-boundary.md` (#199): a match tags Ronan Connolly and stops the run, no
development
--- name: in-your-face description: Capture a chat-only answer into a durable artifact (markdown + HTML, PDF when cheap) and launch it automatically so the user cannot miss it. Use when user says "in your face", "don't let me lose this", "save that answer", "make that durable", or right after answering a substantive side question (a recipe, comparison, how-to, or generated prompt) that would otherwise die with the context. category: workflow argument-hint: [--no-open] [--vault <short>] [hint of
tools
One-shot headless OpenAI Codex CLI calls for background/admin AI tasks — summaries, classification, extraction, admin glue. The default engine for anything that runs AI constantly in the background (daemon-driven, per-event), because it bills the flat ChatGPT subscription instead of Claude usage or per-token API spend, and it keeps working while Claude is rate-limited. NEVER for coding — coding stays Claude. Use when a skill or daemon needs a cheap always-on AI call, when the user says "use codex", "ask codex", "codex as backup", or when building a background summarizer/classifier into a listener or loop. Reads auth from ~/.codex/auth.json (ChatGPT account, no API key).
research
Turn a warranty rejection, repair quote, or RMA email into a cited decision brief — legal read (NL/EU consumer law), is the part user-serviceable, live part and new-unit prices, repair-vs-DIY-vs-new economics, before-you-send-it checklist, deadlines. Use when the user pastes or screenshots a repair quote, warranty rejection, "not covered" email, onderzoekskosten fee, or asks "should I repair or replace this".