agentleFS
armenrarmenr/modelmux/.claude/skills/orient/SKILL.md

orient

Skill2 starsChanged 4 months agoMITView on GitHub

Read .agent-docs/now/handoff.md and surface current session state. Use at session start, after time away, or when reorienting ("where are we?"). Pulls live git state at invocation, verifies the handoff isn't stale against git reality (code ahead of docs → trust the code), runs a quick state-file health sweep (the folded-in /status check), surfaces obligations-ledger deltas — landed / came due / overdue / dangling — from now/obligations.md when it exists, else the handoff's Obligations section, and on multi-party installs runs the Room-threads resume step (read the handoff's continuity block, refresh from the comms log since the handoff timestamp, apply the re-read rule before any post).

27 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use retro, ui-ux-pro-max and 1 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • AWSSkill

    retro

    Conduct a retrospective on a coding session.

    Also in 3 projects that use this

  • StrayDragonSkill

    ui-ux-pro-max

    UI/UX design intelligence. 50 styles, 21 palettes, 50 font pairings, 20 charts, 8 stacks (React, Next.js, Vue, Svelte, SwiftUI, React Native, Flutter, Tailwind). Actions: plan, build, create, design, implement, review, fix, improve, optimize, enhance, refactor, check UI/UX code. Projects: website, landing page, dashboard, admin panel, e-commerce, SaaS, portfolio, blog, mobile app, .html, .tsx, .vue, .svelte. Elements: button, modal, navbar, sidebar, card, table, form, chart. Styles: glassmorphism, claymorphism, minimalism, brutalism, neumorphism, bento grid, dark mode, responsive, skeuomorphism, flat design. Topics: color palette, accessibility, animation, layout, typography, font pairing, spacing, hover, shadow, gradient.

    Also in 3 projects that use this

  • AWSSkill

    ship

    Finalize dev work on an issue branch — update docs, push, and open a PR.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

9 rules. 0 are common, 4 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 25 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If the session touched runtime-facing code (the app, a service, a CLI), a quick check that the claimed state holds (e.g.

    line 40
  2. Distinctive:

    If any contradiction exists, surface it BEFORE acting on the handoff's "immediate next action."

    line 41
  3. Very specific:

    Do NOT chase or settle here — /orient reads and surfaces.

    line 78
  4. Very specific:

    If the handoff carries a ## Room-threads section (/handoff §8.6 — present only on multi-party installs with ≥1 live inter-agent thread), it …

    line 82
  5. Very specific:

    No Room-threads section → nothing to resume.

    line 96
  6. Very specific:

    Staleness / health flags: if any (from §2 + §3.5) — incl.

    line 107
  7. Distinctive:

    After significant time away from the project

    line 115
  8. Distinctive:

    When the user asks "where are we?"

    line 116
  9. Distinctive:

    After context compaction (if the SessionStart hook didn't auto-orient)

    line 117

“Common” is measured: kept by at least 6 in 10 of the 25 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/orient — Surface current session state from handoff

Single-entry session orientation: read the curated bridge, verify it against reality, and surface a ~12-line brief. Folds in the /status quick-health sweep so there is ONE orientation skill, not two.


Live git state (captured at invocation)

Recent commits:

!git log --oneline -10 2>/dev/null || echo "(no git history)"

Working tree:

!git status --short 2>/dev/null || echo "(not a git repo)"


1. Read the handoff

Read .agent-docs/now/handoff.md.

If missing:

  • Note that the agent context system may not be set up here, or the handoff was never regenerated
  • Recommend running /handoff to create one

2. Check staleness vs reality

  • last-modified field. If > 24h old, flag prominently.
  • The live git state above (captured at invocation) — verify it matches the handoff's claims.
  • If the session touched runtime-facing code (the app, a service, a CLI), a quick check that the claimed state holds (e.g. the last bun run build / bun test test/ baseline the handoff cites is still current).
  • If any contradiction exists, surface it BEFORE acting on the handoff's "immediate next action."

Trust the code. When code and docs disagree — the tree/commits above show work the handoff never mentions, or a step the handoff claims is pending already exists on disk — the docs are STALE, not the code. Trust the code, say so explicitly in the brief, and offer to run /handoff to true the docs up BEFORE acting on the handoff's Immediate Next Action.

3. Read supporting docs (in this order)

  • .agent-docs/now/status.md
  • .agent-docs/now/work-plan.md
  • .agent-docs/now/open-questions.md
  • .agent-docs/now/obligations.md (if it exists; otherwise the handoff's ## Obligations section carries the ledger)

3.5 Quick state-file health sweep (the folded-in /status check)

A fast, no-deep-analysis health pass over the durable state surfaces — flag drift, don't fix it (fixes are /flush//handoff's job):

  • now/* currency — last-modified on status.md / work-plan.md / open-questions.md. CONVENTIONS.md lint rule 12 warns at 7d, fails at 90d (the doc linter reports staleness as a warning where installed); flag anything stale.
  • Latest checkpoint — newest file in .agent-docs/checkpoints/; if a /sitrep is more recent than handoff.md, the forensic state is ahead of the bridge — read it (it holds the dead-ends + in-flight pause point the handoff may have summarized away).
  • Open work-units / blockers — count active WU-NNNN in now/work-plan.md and open OQ-NNN in now/open-questions.md.
  • Index lint — if the doc linter hook is installed (Standard profile) and it is cheap, note whether python3 .claude/hooks/lint-docs.py --root .agent-docs is clean; do NOT fix here.

3.6 Obligations deltas (ADR-0012)

The obligations ledger is the inbound/outbound inter-party debt record; at cold-start what matters is what MOVED while you were away. Read it — if now/obligations.md exists, that file; otherwise the handoff's ## Obligations section — and compute the deltas against the git/time gap since the handoff was written.

Where to look — SAME sweep, same "what moved?" question, different surface:

  • If now/obligations.md exists: scan the repo's agent-comms log (tool-generic — whatever coordination log the layer writes) for traffic since the last handoff; a row's Source message-id is your lookup key.
  • Otherwise (the handoff ## Obligations section): ask the operator ("anything land while I was away?") and check each row's named sources — a commit now on disk, a PR / issue state, a dated conversation note. No comms log to scan; the row's Source cell is the pointer.
  • Safety is identical either way — surfacing an overdue still fires only the row's recorded default-if-silent, and a gate row (operator / authorization) never carries apply-default by construction.

Then compute:

  • Landed — an owed-to-me receivable whose deliverable arrived (a cited commit/doc now exists on disk, a settled-strikethrough was added, or — file form — a settling message appears in the comms log) → you may be unblocked; say which HARD row cleared.
  • Came due — an owed-by-me debt whose Due/trigger has passed → you owe delivery now.
  • Overdue — an owed-to-me receivable past its Trigger/by-when with no settlement → apply its default-if-silent: chase-once → send the one ping; apply-default → proceed on the recorded fallback; never-chase-never-peek → record the disposition and move on. Name which default fired. (A gate row — operator / authorization — never carries apply-default by construction, so surfacing an overdue never auto-crosses a gate.)
  • Dangling — an owed-to-me row whose cited OQ-/REV- has since been resolved or removed by another path → the wait may already be moot; flag it "target closed — reconcile" rather than re-chase a settled thing.
  • Rot risk — any owed-to-me row with no parseable Trigger/by-when → flag it "cannot come due, will not be chased."

Do NOT chase or settle here — /orient reads and surfaces; /flush and /handoff mutate.

3.7 Room-thread resume (multi-party installs only)

If the handoff carries a ## Room-threads section (/handoff §8.6 — present only on multi-party installs with ≥1 live inter-agent thread), it is the amnesia-proofing for in-flight conversations on the shared comms layer. BEFORE touching that layer:

  1. Read the section first — thread + subject, participants + locked roles, YOUR role and its status, standing hooks, durable-base pointers, do-not-relitigate markers.
  2. Refresh from the comms log SINCE the handoff's timestamp — the section's message-id anchors are the lookup keys. Threads move faster than handoffs regenerate; the section orients you, the log is ground truth.
  3. Apply the section's re-read rule before posting ANYTHING into a thread: re-read your own last post and all replies to it first. A post made from the handoff's frame alone — without the log refresh — is the confident-wrong failure mode this machinery exists to kill.

No Room-threads section → nothing to resume; skip. Do not post from this step — /orient reads and surfaces; responding to what moved is the session's work, not orientation's.

4. Surface to user (concise — ~12 lines)

  • TL;DR: 1-2 sentences pulled from status.md
  • Immediate Next Action: verbatim from handoff.md
  • Active open questions that gate forward motion (typically 1-3 OQ-NNN from open-questions.md)
  • Active work-unit(s): the WU-NNNN in flight
  • Obligations deltas (from §3.6): one line — what landed (unblocked), what came due (you owe now), what's overdue (which default-if-silent fired), any dangling / rot-risk row. Omit if the ledger is absent or unchanged.
  • Room-threads (multi-party, from §3.7): one line per live thread — thread · your role · what moved in the comms log since the handoff · your next deliverable/hook. Omit when the section is absent.
  • Staleness / health flags if any (from §2 + §3.5) — incl. "a /sitrep post-dates the handoff" if so
  • Recent commits: 3-5 lines from the live git state above

End with: "Ready to proceed with the immediate next action, or refresh state first via /handoff or /flush?"

When to invoke

  • First action in a new session (manual or via SessionStart hook injection)
  • After significant time away from the project
  • When the user asks "where are we?"
  • After context compaction (if the SessionStart hook didn't auto-orient)

When NOT to invoke

  • Mid-operation. If you're working, you already know where you are.
  • Repeatedly within the same session. Read once, work from memory.

Design rationale

The session-lifecycle contract lives in standing-rules-core.md (§Context lifecycle) and CONVENTIONS.md (§6 — the checkpoint contract /sitrep writes and /handoff consumes). Orientation is one skill, not two: the standalone quick-health /status check is folded into §3.5 above. Live git state is pulled at invocation via the shell-interpolation blocks at the top, so the §2 staleness check compares the handoff against current reality rather than a snapshot — and the trust-the-code rule gives that comparison a precedence: the working tree is ground truth, the docs are the map.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.