메뉴
Interpret Alexandria event log wake payloads from the local web UI/runtime, decide whether action is needed, and use AX commands to inspect or write event-log state.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Start Alexandria in the current project. Initializes the project when needed, or welcomes the user back when Alexandria is already configured.
Raven's procedure for running EL3 Front-of-House Walk: launch detached against an EL2 draft bundle, walk the director through section-comprehension checks and held-back problem rulings, bank the director's answer as a Ledger event, and let AX apply only validated section/shape bundle corrections.
Use when writing or revising a library context's LEAD card body — the highest-altitude card whose how-it-does-it story the Index renders for the whole context. The lead must account for every other card in the context: narrate and link its direct children, and `relegates:` its deep internals, so the Index reads as a story instead of a pile of unused words. Applies where card bodies are authored (EL5/atomize), using the relationships Front-of-House has settled.
The authoring standard for product-library card bodies: write the WHAT / WHERE / HOW in the product's own plain language, keep every file path and code identifier in source_evidence (never in the prose), and keep event and ruling ids in frontmatter. Applied wherever a card body is written or revised — the Back-of-House walk's emit_bundle move, Front-of-House corrections, and EL5 body-fill.
Raven's procedure for running EL4 Empty Library Confirm: open a post-EL3 draft bundle as a body-free catalog, guide the director's structural review, and record either a user-authored library.confirmed Ledger event or a structure-only rejection edit list routed back to EL3.
Raven's AX procedure for drafting exactly one Raven Vision onboarding slot from Alexandria source items, using the same per-slot Vision guidance and examples as the canvas-library-spike prototype.
| name | alexandria-event-log |
| description | Interpret Alexandria event log wake payloads from the local web UI/runtime, decide whether action is needed, and use AX commands to inspect or write event-log state. |
Use this skill when a monitor injects an Alexandria event log update into the session.
The injected payload is a wake signal, not a complete task brief. It includes a ledger event with:
id: event id in the Alexandria event logtype: event type, such as canvas.review.requested or
play.startedat: event timestampactor: who or what emitted the eventpayload: event-specific dataRead the injected event object first.
Decide whether the event asks for agent action.
Inspect projected state only when the event does not include enough context:
ax inspect state --json
Use recent raw events only for audit/debug or correlation:
ax inspect events list --json --limit 20
Do not edit events.jsonl, cursor files, connection leases, or subscription
files directly. Use AX commands so validation, idempotency, and projection
stay consistent.
Use ax inspect events schema --json to inspect the supported event types and
payload shapes programmatically.
canvas.review.requested: inspect the referenced canvas/step context, perform
the requested review, then write back a useful result when appropriate.play.requested: the director asked to start a play (e.g. clicked Frame a
Problem on the coin). For frame-the-problem, use the frame-the-problem
skill: confirm you're connected, elicit the material, and launch the play.play.human_input_requested: a running play suspended at a human gate and
needs the director. For frame-the-problem, use the frame-the-problem skill
to read the draft, riff with the director, then answer the exact
(fabroRunId, questionId) with ax raven answer once he agrees. Never
self-answer a gate. For front-of-house-walk, use the
front-of-house-walk skill: record Raven's presented turn, riff the current
Stage-2 question or Hot Spot with the director, and answer with
ax raven answer --bundle <bundle>.play.human_input_resolved: a pending gate was answered (by you or elsewhere).
Clear that ask from your tracked set; do not re-answer it. This is not a
request to start a new play.play.started, play.completed, play.failed, and play.status_observed:
play run lifecycle events. Inspect them for status/debug context; do not treat
a started/completed/failed echo as a request to start another play. For a
frame-the-problem play.completed, use the frame-the-problem skill to
present the finished framing for the director to ratify or loop (no auto-bank).canvas.step.saved: context was saved. Do not wake or respond unless another
event asks for action.raven.vision.source_attached: source context changed. Do not draft from
this event. Acknowledge briefly, for example: "I see you added a source. Let
me know when you're ready for me to draft The Shift."raven.vision.drafting_requested: the director explicitly asked Raven to
begin drafting from the attached Vision sources. Use raven-vision-drafting
to draft exactly one current Vision slot if no slot is already waiting for
review.raven.vision.slot.updated: a Vision slot draft changed. Inspect projected
state before acting; if the actor is Raven, wait for user review rather than
writing another slot. If the actor is the Viewer/user, preserve Raven's notes
and wait for approval or an explicit revision request.raven.vision.slot.approved and raven.vision.slot.skipped: user review
feedback. Use raven-vision-drafting to draft the next empty Vision slot,
then stop for Viewer review.raven.source_of_truth.updated: Raven's internal Source of Truth document was
written or refreshed. Inspect raven.sourceOfTruth in projected state before
using it as durable context.raven.vision.banked: Vision is banked into Raven's Knowledge Bank. Treat the
banked Source of Truth as context only; do not generate Library cards from this
event.library.front_of_house.turn_recorded,
library.front_of_house.answer_recorded,
library.front_of_house.bundle_patch_applied, and
library.front_of_house.residual_gap_recorded: EL3 audit/provenance events.
Use them to verify director-answer provenance and residual-gap accounting. A
process-authored residual whose reason starts with settled by triage: is a
machine-made triage settlement, distinct from director-ruled answers and
frame-ruling cascade settlements.library.front_of_house.item_reopened: EL3 audit event saying a
triage-settled agenda item was reopened for a later stage-next. It does not
delete the original settlement; inspect both events when auditing the item.
This is not a wake request and should not start another play.library.confirmed: EL4 durable approval. Treat it as the only approval fact
when actor.kind = user and product, bundle path, and library version match.
Do not create a second approval flag or broaden the approval to a later
version.library.confirmation_rejected: EL4 rejection/edit list. This is not an
approval. Use the empty-library-confirm skill to interpret it, then route
the edit list back to front-of-house-walk.When writing an agent result to the event log, use
ax inspect events append. Run ax inspect events append --help to see the
supported event types and append options.
For Raven Vision slot collaboration, prefer the narrower runtime command over a generic event append:
ax raven vision slot update --slot <slot-id> --text "<draft text>" --notes "<markdown notes>" --json
ax raven vision slot approve --slot <slot-id> --json
ax raven vision slot skip --slot <slot-id> --json
Do not create temporary files just to pass Raven Vision draft text or notes. This command writes the validated state event through AX.
Use update for one Raven-authored draft with user-visible notes, then stop
until the director edits, approves, or skips it. Raven must not self-approve.
If the director explicitly approves or skips in Claude Code, record that
decision with approve or skip instead of ax inspect events append; after
the command succeeds, continue only from the projected state and the current
drafting trigger rules.