runway
Skill2 starsChanged 33 days agoMITView on GitHub
Ships the bragDoc's public channel to the user's own website (a static site in its own git repo, next to cockpit). Commands - publish, promote, ship, showcase, status, rebuild, live-check, profile, release current-employer; also waits for a new post to reach the site. Use when the user talks about the website, the portfolio, publishing or previewing site content, case studies on the homepage, the site's resume PDF, or deploying.
6 other projects use a version of this.
Adopted by
Most recently taken up by
- s3ntin3l8/runway-ai-usage-trackertheir own version, first seen Nov 8, 2026
- JustinPerea/nebula-nodestheir own version, first seen Nov 5, 2026
- udaysharmadev/Skillstheir own version, first seen Sep 17, 2026
- WellApp-ai/skillstheir own version, first seen Sep 10, 2026
- ozkilim/runwaytheir own version, first seen Aug 21, 2026
- runapi-ai/runwaytheir own version, first seen Jul 19, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- JustinPerea/nebula-nodes★ 23same name
- udaysharmadev/Skills★ 13same name
- WellApp-ai/skills★ 9same name
- s3ntin3l8/runway-ai-usage-tracker★ 3same name
- ozkilim/runway★ 2same name
- runapi-ai/runway★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
4 rules. 0 are common, 0 are distinctive, and 4 are very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 57
After every command that changes files: validate, commit runway: <command>: <summary> in cockpit, push.
- Very specific: line 64
Publishing: every promote (content or code) runs runway.py profile --write: git pull --rebase, replace the block, commit, push.
- Very specific: line 81
After I approve: runway.py promote --initial (marks main as published).
- Very specific: line 84
Every RUNWAY command ends by listing the open items of any checklist in data/extrovert/checklists/ (see EXTROVERT checklist), one line each.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
RUNWAY
No command given (bare /runway): read COMMANDS.md and print its ## RUNWAY section verbatim (heading through the line before the next ## ) in a fenced code block. Nothing else, no commit.
RUNWAY moves the bragDoc's public channel into the site repo, one way only. python scripts/runway.py does all of it: building the data, every gate, git in the site repo, hashes. RUNWAY generates nothing: renovate already wrote every summary, card, slug and year (see the bragdoc skill). If anything is missing or out of date, publish blocks and names it; fix it through renovate, never at publish time.
- Config:
data/runway/config.json(site path, remote, branches, build/check/test commands, preview-URL pattern, deploy hook, field map, impact bands, PDF name, the standing visibility default; every key is inscripts/schemas.md). - State:
data/runway/state.json(showcase, showcase intents, retired slugs, email confirmation, last publish/promote, shipped code branches). - The site: any static site that reads one JSON file (the site data contract in
scripts/schemas.md), in its own repo atconfig.site_repo(a sibling of cockpit, never inside it), hosted anywhere that deploysmainto production and gives other branches a preview URL. - Its git: branch
mainplus acontentbranch for previews (deleted after each promote, recreated by the next publish), the same identity rules as cockpit. Never force push. - What publish writes: only
config.data_pathand the public resume PDF. Any other site file (components, styles, tests, workflows) changes only on my explicit instruction: make the change in the site working tree onmain, uncommitted, then ship it (below). Never commit site code straight tomain.
What reaches the site
| Site field | From |
| entries | entries with visibility: "public" only |
id | site_slug: cut once from the public headline, never changed (deep links /bragdoc#<id> must never break); a retired slug is never reused |
field | category → config.category_to_field |
impact | rank_score → config.impact_bands (e.g. 8–10 → 3, 5–7 → 2, 1–4 → 1) |
year | entry year (period, else data/bragdoc/project_years.json, else a single-year tenure) |
title / summary / stack | public_summary headline / summary / stack |
metric / context / did / result / where | site_card |
highlight | state.showcase (max 3; public; never the current employer's), with showcase intents applied |
pdf | the full public bragDoc (render.py public-resume, header names.public_full), served at /<pdf.file> |
Gates (every publish; any failure writes nothing)
- Contract check: field ids exist, impact 1–3, ≤ 3 highlights, unique slug ids, every required string present.
- Disclosure:
disclosure_lint --channel publicon every string, per entry (with the entry's employer: relative cost figures only from past employers, clients or personal work; never the current employer; absolute amounts never), and again as one document. - The PDF: its HTML, its extracted text and metadata pass the public channel, plus the local-path and username patterns (fourthwall deny list
regex+local_identity, anyfile://). - The site's build (
config.build_cmd): files it rewrites (config.synced_files, e.g. a hosting config whose CSP hashes follow the pages) are committed with the data. - The site's copy:
voice_lint.py --site <dist>on the fresh build (the voice rules' em-dash limit on titles, meta and link-preview text, visible text, alt and aria-label; text markeddata-source, from another platform, only warns). - The site's check and tests:
config.check_cmd, thenconfig.test_cmdagainst the fresh build served ontest_port(for example smoke and accessibility tests).
Commands
| Command | Does |
/runway publish | runway.py publish: gates → content branch (merged up from main) → write data + PDF → build + check + tests → commit → push → print the preview URL. Nothing is live. Show me the URL and wait |
/runway promote | Only after I approve the preview: runway.py promote merges content → main, pushes, and sets published_hash (from bragdoc_public.py payload-hash) for every published entry; entries made private since get published_hash: null. Refuses if the bragDoc changed after the publish. Then it waits for the production deploy and runs the live tests (see "After every promote and rebuild") |
/runway ship "<summary>" | runway.py ship "<summary>" [--slug <slug>]: for site code changes I asked for, sitting uncommitted on main. Refuses content files (they go through publish). Runs the gates on main + the changes first (nothing is created if one fails), then creates code/<slug> (≤ config.code_slug_max characters, so the preview hostname works), commits (with config.commit_trailer), pushes, returns the site repo to clean main, records state.ships, and prints the preview URL. Show me the URL and wait |
/runway promote code/<slug> | Only after I approve that preview: runway.py promote code/<slug>: if main moved since the ship, merges it into the branch and runs the gates again; then fast-forwards main, pushes, waits for the production deploy and runs the live tests |
/runway prune | runway.py prune: deletes every site branch fully merged into main (local and origin) and lists any other branch with its commits, which stays until you've seen it. Promotes already delete their own branch; the next publish recreates content from main, so a preview exists only while something waits for your approval. Old preview deployments stay in your host's dashboard until you delete them there |
/runway showcase <ids> | runway.py showcase <entry_ids> (max 3; public; never current-employer). runway.py showcase intent <staging_id> --replaces <entry_id>: once that staging item is probed, renovated and public, its entry takes the slot at the next publish |
/runway status | runway.py status: live vs pending (bragdoc_public.py pending), gate problems, showcase intents, and launch blockers (the site email not set or not yet confirmed, the PDF not on main, no preview URL). Then Rebuilds and production: the local deploy hook (present / valid / missing, never shown), whether a scheduled rebuild is verified, when production was last built, the posts feed's health, and the last live check |
/runway rebuild | runway.py rebuild: POSTs the deploy hook from config.deploy_hook_file (gitignored; validate.py fails if it's ever tracked), then waits for the production deploy and runs the live tests. If the file doesn't exist (e.g. a cloud session), tell me to trigger the site repo's scheduled-rebuild workflow by hand |
| Post confirmed (EXTROVERT) | runway.py await-post <post-url>, run in the background: polls the blog feed (config.posts.feed: dev.to, Medium or any RSS feed) until that post is in it, then rebuilds. On timeout it rebuilds anyway and says the scheduled rebuild will catch a slow feed. Report the result when it finishes |
| "verify the daily rebuild" | runway.py verify-action in the background, then tell me to run the site repo's rebuild workflow by hand: a production redeploy proves its secret works. Records rebuild_action_verified. No click: runway.py verify-action --scheduled in the background waits for the next scheduled run, then checks that production redeployed after it (90 min grace); trigger no other deploy in that window |
/runway profile | runway.py profile: print the cockpit block with its gates and whether the profile README is current; runway.py profile --write updates it (normally done by promote) |
/runway live-check | runway.py live-check: the live tests against production now, no waiting |
| "release current-employer entries" | My one instruction: runway.py release current-employer makes every private current-employer entry public, then publishes (content branch + preview only) |
| "the test email arrived" | runway.py confirm email (the email launch blocker clears only on my word) |
After every command that changes files: validate, commit runway: <command>: <summary> in cockpit, push.
A profile README (optional, config.profile)
If you keep a profile README (for example a GitHub profile repo), cockpit can own one block of it: the text between <!-- COCKPIT:START --> and <!-- COCKPIT:END -->. Everything else stays yours, byte for byte (line endings too), and any other generated block (such as a latest-posts list written by that repo's own workflow) is never touched. Never edit outside the markers without my explicit instruction.
- The block (
runway.py profileprints it): the two-line intro fromconfig.profile.intro; Selected work: the topconfig.profile.top(5) public entries, each its public headline linked to<site_url>/bragdoc#<slug>, plus the metric only when the headline lacks its numbers; then links (website, Medium, LinkedIn) fromprofile.jsoncontacts on the public channel. - Selection: highlighted entries first; then impact; then recency (year); then
rank_score; a metric without a number only breaks the remaining ties; at mostconfig.profile.per_project(2) per project (the first part of the card'swhere). Impact comes fromrank_score(my 1–10 assessment at renovate, never re-rated without asking) throughconfig.impact_bands. - Gates: each entry's title and metric pass
disclosure_lint --channel publicwith that entry as context, the rest as one document; the visible text passes the EXTROVERT voice lint. Any failure writes nothing. - Publishing: every promote (content or code) runs
runway.py profile --write:git pull --rebase, replace the block, commit, push; nothing happens when the block is unchanged; a failure is reported and never undoes the site promote.
After every promote and rebuild
runway.py reads production's Last-Modified before the trigger, waits (config.deploy_wait_s, 15 min by default) until production serves a newer deployment, warms the cache (config.warm_paths), then runs config.live_test_cmd against production and records the result in state.last_live_check. Report: when it deployed, passed/failed, and any failing test. If production doesn't redeploy in time, say so: a failed build usually keeps the previous version live, so check the deployment on the host. These waits are long: run promote, rebuild and await-post in the background and report when they finish. --no-wait skips the wait and the tests.
The deploy hook is a secret
Never print, log, echo or paste the hook URL (not even part of it), and never put it in a commit, a message or an error. runway.py validates it against config.deploy_hook_pattern and reports only present / valid / missing; request failures report only the HTTP status or the error type. A copy for a scheduled rebuild lives only in the site repo's CI secrets; that workflow should run with no extra permissions, fail with a clear error if the secret is missing, and never print it.
Standing visibility default
New entries get config.visibility_default (the template: private; for example "public unless they belong to the current employer"). It is the user's own rule, applied once, by bragdoc_public.py add-entry at renovate. The current employer is always the open role in employment.json: a job change re-tiers the cost and highlight rules automatically but never flips an existing entry's visibility. An employer whose NDA makes its public restrictions permanent (data/bragdoc/nda.json) stays under every current-employer rule after the job ends, and release current-employer never releases its entries.
After renovate (automatic)
Renovate ends with runway.py publish --if-pending: when any public payload (or the site data) changed, it publishes to the content branch and prints the preview URL; otherwise it does nothing. Promote is never automatic: it is the one human gate before anything goes live.
Launch (first run)
runway.py previewwrites the data + PDF into the site working tree (uncommitted), builds and checks. Show merunway.py tableand the PDF; wait for my OK.runway.py publish --initial: the very first commit, straight onto the empty repo'smain, push. Stop.- After I approve:
runway.py promote --initial(marks main as published). I connect the repo to a host and point the domain at it; setconfig.preview_urlonce the host builds branches.
Open checklists
Every RUNWAY command ends by listing the open items of any checklist in data/extrovert/checklists/ (see EXTROVERT checklist), one line each.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

