Menu
Screen a Target against the hard gates (the screen verb). Applies auto — the write IS the screen.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
The market-radar's sensing pass — sweep recent public signal (HN + Reddit, free, recency-windowed) on a company or theme and bucket every hit against the contract's watch signals (distress, ownership, momentum, valuation, AI-pressure). Use to scout a candidate or refresh the pipeline.
Run the model's audit (the model's audit service, mechanical subset) and summarize — green/red, defects with the invariant each violates. Run after any batch of edits.
Close an acquisition: create the Deal + Business + Interface(s) (GATE 2: go/no-go is the user's).
The desk over the whole deal machine — pipeline status, pending proposals, and which verb applies next. Start here ("/deal-desk") to see the state and be guided to the right action skill.
The hiring assistant — run a candidate through Bending Spoons' selection funnel and score them against the disclosed gates. Read-only, advisory: it drafts a recommendation; the committee decides. The people-side twin of screen-target.
Create a new Target in the pipeline (the source verb). Human-gated: admission needs the user's approval.
| name | screen |
| description | Screen a Target against the hard gates (the screen verb). Applies auto — the write IS the screen. |
| disable-model-invocation | true |
Execute the screen verb of the Bending Spoons world model, per its contract.
capabilities/actions/screen-target.md — parameters, submission criteria, gate. Follow it exactly.python3 .claude/skills/radar-sweep/scripts/radar_sweep.py "TKR" --gates
(revenue band · HQ · sector, [derived] from EDGAR)... "Legal Name" --company (GLEIF → ESEF chain; honest n.d. for privates)--loyalty "Brand" as the predictable-earnings proxy
(rating base · interest trend · community)
Fill the nine gate values from these, marked by provenance; ask the user only for what remains.python3 -c "import sys; sys.path.insert(0,'mcp'); import engine, json; print(json.dumps(engine.propose('screen', PARAMS), indent=2))"
If the engine REFUSES, report the refusal verbatim — it is a submission criterion firing (a will
constraint or an invariant), not an error to work around.python3 -c "import sys; sys.path.insert(0,'mcp'); import engine; print(engine.apply('PROPOSAL_ID', 'USER_NAME'))"
The engine runs the audit and commits if green; report the outcome. Never edit the csvs directly.