Look

Convert user images into validated action drafts like calendar events, meal macros, receipts, products, civic reports, or documents for confirmation and exec...

Look bridges the physical world and the digital agent — it takes a user-provided image, infers what the user probably wants done with it, and produces a validated, decision-ready action draft across domains including calendar events, meal macros, places, product comparisons, receipts, documents, and civic reports. It resolves ambiguity through research and option reduction before asking any clarifying questions, and nothing executes without explicit per-draft confirmation.

When to use

• A flyer or poster → calendar event or ticket purchase draft
• A meal photo → macro estimation
• A storefront or sign → save to try-list
• A product photo → pricing comparison or order draft
• A civic issue photo → 311 report draft
• A receipt → expense entry
• A document → searchable PDF and filing draft

What this skill does not do

• Generic OCR utility
• Universal computer vision toolkit
• Background surveillance or tracking
• Generic automation framework

Responsibility boundary

Look owns image-to-action conversion: ingest, context inference, domain routing, draft generation, and execution with confirmation.

Look does not own: web research (Sift), knowledge graph writes (Elephas), preference persistence (Taste), communications (Dispatch).

Supported domains

Events, food macros, places, products, civic issues, receipts, documents.

Read references/domain_playbooks.md for detailed per-domain behavior.

Core workflow

1. Ingest image(s) with EXIF if available
2. Merge extraction evidence (OCR, entities, context)
3. Infer context and likely intent
4. Route to appropriate domain
5. Research and validate externally
6. Filter by constraints (dietary, preferences, permissions)
7. Reduce options before asking questions
8. Generate 1-3 decision-ready ActionDrafts
9. Emit Signal files for extracted entities (places, events, products) to ~/openclaw/db/ocas-elephas/intake/{signal_id}.signal.json. Use Signal schema from spec-ocas-shared-schemas.md.
10. Write journal via look.journal

Clarification happens only after option reduction, not before.

Commands

• look.ingest.image — ingest image(s) with optional EXIF and device pre-parse
• look.propose.actions — generate ActionDrafts with DecisionRecords
• look.execute.action — execute a confirmed draft (requires explicit approval)
• look.rollback.action — attempt rollback for reversible actions
• look.status — last ingest, pending drafts, items awaiting confirmation
• look.config.set — update configuration
• look.journal — write journal for the current run; called at end of every run
• look.update — pull latest from GitHub source; preserves journals and data

Confirmation and rollback rules

• Draft-first always. No execution without explicit confirmation.
• High-risk actions (purchases, 311 submission, health writes): require per-draft confirmation token.
• Reversible actions (calendar, maps): expose rollback information.

Permission discipline

Default deny. Request minimally. Drafting continues even without execution permissions. Blocked execution reported, not silently skipped.

Boundaries

• Never invent OCR text, barcodes, prices, or license plates
• EXIF capture location is not the event venue
• iOS relay pre-parse is optional evidence, not truth
• The skill must work without relay upload

Storage layout

text

~/openclaw/data/ocas-look/
  config.json
  state.json
  events.jsonl
  decisions.jsonl
  reports/
  artifacts/

~/openclaw/journals/ocas-look/
  YYYY-MM-DD/
    {run_id}.json

Default config.json:

json

{
  "skill_id": "ocas-look",
  "skill_version": "2.3.0",
  "config_version": "1",
  "created_at": "",
  "updated_at": "",
  "domains": {
    "events": true,
    "food": true,
    "places": true,
    "products": true,
    "civic": true,
    "receipts": true,
    "documents": true
  },
  "user_profile": {
    "diet": "vegetarian"
  },
  "commerce": {
    "auto_purchase": false
  },
  "retention": {
    "days": 30,
    "max_records": 10000
  }
}

OKRs

Universal OKRs from spec-ocas-journal.md apply to all runs.

yaml

skill_okrs:
  - name: draft_accuracy
    metric: fraction of ActionDrafts accepted without modification
    direction: maximize
    target: 0.75
    evaluation_window: 30_runs
  - name: domain_routing_accuracy
    metric: fraction of images routed to correct domain on first attempt
    direction: maximize
    target: 0.90
    evaluation_window: 30_runs
  - name: confirmation_compliance
    metric: fraction of high-risk actions requiring confirmation before execution
    direction: maximize
    target: 1.0
    evaluation_window: 30_runs

Optional skill cooperation

• Sift — web research for validation during draft generation
• Elephas — emit Signal files for extracted entities after draft generation

Journal outputs

Observation Journal — all image ingestion and draft generation runs.

Initialization

On first invocation of any Look command, run look.init:

1. Create ~/openclaw/data/ocas-look/ and subdirectories (reports/, artifacts/)
2. Write default config.json and state.json if absent
3. Create empty JSONL files: events.jsonl, decisions.jsonl
4. Create ~/openclaw/journals/ocas-look/
5. Ensure ~/openclaw/db/ocas-elephas/intake/ exists (create if missing)
6. Register cron job look:update if not already present (check openclaw cron list first)
7. Log initialization as a DecisionRecord in decisions.jsonl

Background tasks

Job name	Mechanism	Schedule	Command
look:update	cron	0 0 * * * (midnight daily)	look.update

text

openclaw cron add --name look:update --schedule "0 0 * * *" --command "look.update" --sessionTarget isolated --lightContext true --timezone America/Los_Angeles

Self-update

look.update pulls the latest package from the source: URL in this file's frontmatter. Runs silently — no output unless the version changed or an error occurred.

1. Read source: from frontmatter → extract {owner}/{repo} from URL
2. Read local version from skill.json
3. Fetch remote version: gh api "repos/{owner}/{repo}/contents/skill.json" --jq '.content' | base64 -d | python3 -c "import sys,json;print(json.load(sys.stdin)['version'])"
4. If remote version equals local version → stop silently
5. Download and install:

   bash

   TMPDIR=$(mktemp -d)
   gh api "repos/{owner}/{repo}/tarball/main" > "$TMPDIR/archive.tar.gz"
   mkdir "$TMPDIR/extracted"
   tar xzf "$TMPDIR/archive.tar.gz" -C "$TMPDIR/extracted" --strip-components=1
   cp -R "$TMPDIR/extracted/"* ./
   rm -rf "$TMPDIR"
   

6. On failure → retry once. If second attempt fails, report the error and stop.
7. Output exactly: I updated Look from version {old} to {new}

Visibility

public

Support file map

File	When to read
references/schemas.md	Before creating evidence, drafts, or receipts
references/domain_playbooks.md	Before domain routing or draft generation
references/decision_policy.md	Before risk assessment or confirmation decisions
references/command_reference.md	Before any command execution
references/storage_and_config.md	Before config changes or storage operations
references/journal.md	Before look.journal; at end of every run