find-all

Find All

Run both finder agents in parallel, then brief the user on what the project's gaps mean. Discovery only — no plans, no execution, and the survey list is the only file it writes.

Step 1: Spawn Both Finders in Parallel

Launch two sub-agents simultaneously:

Task finder — spawn with subagent_type: "ops:task-finder" and this prompt:

Scan depth: full

This is a discovery scan. Do NOT add anything to the backlog, and do NOT write a
follow-up file — return every item you find in your response, with your usual
fields, and let the caller decide what to persist.

Opps finder — spawn with subagent_type: "ops:opps-finder" and this prompt:

Your job: identify the new directions of work genuinely worth exploring
(often 3-5; fewer or none is a valid result).
Stay grounded in the project's goals in charter.md and its limitations in MEMORY.md.
Every suggestion must be something an autonomous agent can start working on.

Step 2: Consolidate and Present

Once both agents complete, brief the user on what the two scans add up to.

Format

No tables. Present 3-6 themes, grouped by what they mean for the project rather than by which finder produced them — e.g. "nobody has checked whether the two biggest funders still take unsolicited applications", "the audience numbers everything else rests on are six months stale". Each theme gets a short paragraph: what is missing or going stale, what it costs to leave it alone, and — where there is a real choice — two directions with what each buys and gives up. Effort labels, lens names, per-item file paths, and the agents' own count lines stay out. Close with the two or three you would start with and why, then ask which to pursue — and where the answer depends on what the user wants the project to achieve, ask that instead of guessing.

Nothing gets dropped: write the complete item list, with the finders' own fields, to plans/YYYY-MM-DD_survey.md — a findings list, not a plan — for whoever plans from it later, and mention that file in one line.