agentleFS
hazarsozerhazarsozer/ren-os/skills/pin/SKILL.md

pin

Skill2 starsChanged 2 months agoMITView on GitHub

Use when the friend wants to reactively pin or correct memory mid-session: "remember it like THIS" or "that's wrong, drop it." Triggers on the /ren:pin slash command. The simplest producer in the system — one invocation, one proposal, queued through the single write-queue like every other producer. Not a pipeline.

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No page and no clear target can be resolved → ask which page before proposing anything (never guess and DELETE)

    line 54
  2. Very specific:

    Call skills.pin.lib.pin(text, page, session) or skills.pin.lib.correct(page, replacement, session):

    line 59
  3. Very specific:

    Write to a wiki page directly.

    line 82

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

pin

Reactive memory control, mid-session. The friend says "remember it like THIS" or "that's wrong, drop it" — this skill turns that into exactly one Proposal at the data-plane door (Task 2.1). A pin is the friend's own words (writer="human"), so it applies immediately with provenance and a one-step revert — it never writes a wiki page directly, but it doesn't wait on anyone either. Only an instruction-plane target (global/, decisions/, patterns/, research/) or a detected contradicts conflict leaves it pending instead of applying right away — NOTHING is on disk in that case, so never report a held pin as saved: surface the queue id and tell the friend the pin is pending their approval (/ren approval path: lib.memory.queue.approve_and_apply(<qid>, who)).

When to use this skill

  • Friend invokes /ren:pin "<text>" [--page <path>] — pin new or updated content to a page. If --page is omitted, the caller resolves a sensible default (e.g. the active project's current-context page) before calling pin().
  • Friend invokes /ren:pin --wrong <page> [--instead "<text>"] — a correction. With --instead, the page is replaced with the given text (UPDATE). Without it, the page is proposed for deletion (DELETE) — "that's wrong, just drop it."
  • Friend says: "remember this exactly", "no, that's wrong", "pin this", "correct that page" — confirm the target page once if ambiguous, then call the lib function.

When NOT to use this skill

  • Friend wants to look something up → /ren:recall <query>, not /ren:pin
  • Friend wants the whole session consolidated → /ren:wrap
  • No page and no clear target can be resolved → ask which page before proposing anything (never guess and DELETE)

Behavior

  1. Resolve the target page (wiki-relative path) and the active session id.
  2. Call skills.pin.lib.pin(text, page, session) or skills.pin.lib.correct(page, replacement, session):
    • pin proposes ADD if the page doesn't exist yet, UPDATE if it does.
    • correct proposes DELETE when no replacement text is given, else UPDATE.
    • Both always set producer="pin", writer="human", salience=True.
  3. Both calls go through lib.memory.queue.propose_and_apply, which scrubs secrets, dedups against existing pending entries, and runs conflict detection exactly like any other producer — pin gets no special exemption. Per the v2.2 pivot, a pin/correction to a data-plane page auto-applies immediately (provenance-tagged, one-step revertible). An instruction-plane target (global/, decisions/, patterns/, research/) or a contradicts conflict holds the entry pending instead — nothing lands on disk.
  4. Confirm to the user, honestly per outcome:
    • Applied: Saved (write <write_id>) — <op> <page> (pin), or ... (correction), and mention saying "undo <write_id>" to revert — matching ingest's conversational closing copy.
    • Held pending (instruction-plane target or contradiction): NEVER say "saved". Report Queued as <qid> — <page> is an instruction-plane target, held for your approval and point at the approval path (lib.memory.queue.approve_and_apply("<qid>", who="<friend>"), i.e. the /ren approval flow) or contradiction resolution as appropriate. Exception: when the friend has ALREADY approved the correction in conversation (e.g. wrap's "Live pins" gate asked and they said "delete"), pass approved_by=<friend's handle> to correct(...) — it completes the instruction-plane hold through approve_and_apply in the same call, so the confirmed correction actually lands (dogfood-2 finding H1). A contradicts hold is never auto-completed this way — it still comes back pending and must be reported as held.

Why salience=True

Per spec §3.2, a pinned or corrected page is something the friend explicitly cared enough about to interrupt flow for. salience is carried through the Proposal into the QueueEntry so wake-up's relevance ranking (Phase 5) can boost it — a pin is a strong, direct signal, stronger than anything inferred.

Task-shaped pins (issue #25)

A pin whose content is a TASK ("next session, do X") should carry its own disposal as its last step — end the pinned text with "…and delete this page once done." Belt-and-braces: /ren:wrap's "Live pins" gate (see skills/wrap/SKILL.md) also lists every live pin at session end and asks the friend about ones that look acted-on, so a finished plan-pin gets caught there even when the text forgot its own delete step. Disposal is always the normal correction path (correct(page, None, session) → queue DELETE) — never a direct file deletion.

Standing rules scoped to a project (#63)

When the friend asks to make something a standing rule for the CURRENT PROJECT specifically ("always do X in this repo", "make that a standing rule here") — not an ordinary page pin — the session calls lib.memory.promotion.promote_to_project(text, slug, session) instead of pin(). It proposes an ADD/UPDATE against projects/<slug>/instructions.md (born on first promotion — no skeleton stamping anywhere else) and, being an instruction-plane page (Task 5), always comes back pending. After the friend confirms verbally, complete the hold with queue.approve_and_apply(qid, who="human:pin") — same completion contract as correct(..., approved_by=...)'s _complete_if_held. pin() itself stays unchanged for ordinary page pins; this is a separate entry path, not a mode of pin().

What this skill does NOT do

  • Write to a wiki page directly. Every pin/correction is a Proposal at the data-plane door (propose_and_apply) — applying (or holding on a contradiction) is owned by the queue/tier machinery, not this skill.
  • Run any pipeline, retry loop, or multi-step flow. One invocation → one proposal. If that proposal conflicts with something, queue.propose_and_apply attaches the conflict; a contradicts conflict is held for the model to reason about and resolve (recording that reasoning), with the friend asked only on genuine ambiguity — this skill doesn't try to resolve it itself.
  • Maintain its own state file, hot tier, or session-notes equivalent. There is no --instinct mode in 0.2 (donor skills/note's instincts hot tier is out of scope here — see the harvest map).
  • Decide apply policy. Whether a pin's entry applies immediately or is held pending (an instruction-plane target — global/, decisions/, patterns/, research/ — or a detected contradicts conflict) is governed by the memory write path, not by this skill.

Failure-degradation modes

FailureBehaviorUser-visible
Empty text on /ren:pinRefuse, prompt for text"What should I pin? Usage: /ren:pin "<text>""
--wrong with no resolvable pageRefuse; nothing proposed"Which page is wrong? Usage: /ren:pin --wrong <page>"
Content contains a detected secretpropose_and_apply() raises SecretsFound; nothing queued"Refused: that text looks like it contains a secret."
Duplicate pending pin (same page + same content)propose_and_apply() returns the existing entry (idempotent)"Already queued as <qid>."

References

  • Task 2.1 (lib/memory/queue.py) — the single write-queue this skill's only consumer
  • Spec §3.1 producer 3 — pin/correction verb definition
  • Spec §3.2 — salience and wake-up ranking
  • skills/note/ (donor, ~/Dev/startup-framework) — the pre-0.2 shape this skill shrinks from

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.