# Td Feature Synthesize

> Consolidate the five tdmcp surveyor reports into one prioritized FEATURE_BACKLOG.md — dedupe cross-surface overlaps, reconcile every candidate against the roadmap (NEW vs planned vs extension), rank by a selectable impact×effort weighting profile into P0/P1/P2 (high-confidence first), and emit per-surface tables + a recommended-next shortlist that hands straight to the build pipeline. Use when a…

- **Type:** Skill
- **Install:** `agentstack add skill-hybridlabor-api-bdb-dev-optimized-agent-skills-td-feature-synthesize`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [hybridlabor-api](https://agentstack.voostack.com/s/hybridlabor-api)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** Apache-2.0
- **Upstream author:** [hybridlabor-api](https://github.com/hybridlabor-api)
- **Source:** https://github.com/hybridlabor-api/bdb-dev-optimized-agent-skills/tree/main/mcps/tdmcp/.claude/skills/td-feature-synthesize

## Install

```sh
agentstack add skill-hybridlabor-api-bdb-dev-optimized-agent-skills-td-feature-synthesize
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# td-feature-synthesize — merge, reconcile, prioritize

You consolidate up to five independent surveys into the single artifact the user wants: an honest, prioritized list of new features, organized by surface, ranked by value vs. effort under a selectable weighting profile, and ready to feed the build pipeline. Surveyors brought vetted depth; you bring judgment.

## Procedure

### 1. Gather inputs
Read every `_workspace/discovery/01_survey_{controls,library,cli,ai,td-depth}.md` that exists, plus `docs/ROADMAP.md`, `CLAUDE.md`, `CHANGELOG.md`, and the `project-td-mcp-competitive-landscape` memory if present. If a survey is missing, proceed and record the gap in a coverage line.

### 2. Dedupe
Merge duplicates and near-duplicates into one entry under its best-fit surface, with a `(also raised under )` cross-ref. A control that's really a TD capability, or a CLI command that just exposes a tool, is **one** feature — not two.

### 3. Reconcile against the roadmap (honestly)
Sort every surviving candidate into three buckets and keep all three visible:
- **NEW / unlisted** — the headline value. These are what the user can't already see in the roadmap.
- **Already planned** — cite the phase (Phase 13 / deferred-v0.6.0+). Report them so the user sees what's queued and can re-prioritize, but never relabel them as discoveries.
- **Extensions** — concrete extensions of shipped tools/commands/prompts.

The credibility of the backlog *is* the deliverable — the user owns this repo and will notice inflation.

### 4. Prioritize
Score each item on impact (High/Med/Low) × effort (S/M/L), then assign a single priority:

| | S effort | M effort | L effort |
|---|---|---|---|
| **High impact** | **P0** (quick win) | P0 | P1 |
| **Med impact** | P1 | P1 | P2 |
| **Low impact** | P2 | P2 | **P2 (trap — flag it)** |

**Weighting profile.** How you weight *impact* is the most consequential call you make. Use a named profile — default unless the orchestrator passes one in the run:

| Profile | Leads with | When the user wants… |
|---|---|---|
| **`live-show`** (default) | Does this make a real set better? (audio/beat/camera-reactive, mixing, recovery, hands-free) → **quick-wins (S effort) as the tie-breaker** | the core VJ thesis — the safe default |
| `quick-win` | Lowest effort first; rank by impact÷effort, S before M before L | momentum, a release this week |
| `parity` | Competitor gaps first (8beeeaaat / Embody / dotsimulate LOPs) | closing a perceived feature deficit |
| `agent-dx` | Token-cost / agent-ergonomics wins first (cheap reads, batch, surgical edits) | making the agent cheaper & faster |

Whatever the profile, two rules always hold: a feature that makes a real show better or an agent dramatically cheaper outranks a generic nicety; and within a priority tier, **list higher-`Confidence` items first** (a vetted P1 beats a speculative P1). Respect the project's deliberate broad-tool-surface stance: propose only *additive* features, never consolidation-for-score.

### 5. Emit `FEATURE_BACKLOG.md`
Structure:

```
# tdmcp — Feature Backlog (discovery )

## Executive summary

## Coverage

## Recommended next — Top N

## By surface
### Artist controls & creative tools

### Library, packaging & distribution

### CLI & developer DX

### AI & LLM integration

### TouchDesigner depth (bridge + operators)

### Cross-cutting

## Roadmap alignment

```

### 6. Make it pipeline-ready
Phrase each recommended item so it can be handed straight to `tdmcp-pipeline` (e.g. "Build `create_sdf_text` — Layer 1 GLSL SDF text generator with beat-flash"). Carry any `probe-first` / `UNVERIFIED` flag through so the build harness knows to validate live first.

## Output + return
- Write `_workspace/discovery/FEATURE_BACKLOG.md`.
- Return a tight, self-contained prose summary (the orchestrator relays it): the headline NEW ideas, the recommended shortlist, and any coverage gap.

## Quality bar
- **Decide, don't hedge** — one surface, one priority, one effort per item. If surveys disagreed, pick the more conservative effort and note it in one line.
- **Tables over prose** in the per-surface sections; reserve prose for the summary and roadmap-alignment.
- **No invented operators** survive into the backlog without a `probe-live` flag.
- **Honest buckets** — if most candidates are already on the roadmap, say so plainly; the NEW set's size is a finding, not a target.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [hybridlabor-api](https://github.com/hybridlabor-api)
- **Source:** [hybridlabor-api/bdb-dev-optimized-agent-skills](https://github.com/hybridlabor-api/bdb-dev-optimized-agent-skills)
- **License:** Apache-2.0

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-hybridlabor-api-bdb-dev-optimized-agent-skills-td-feature-synthesize
- Seller: https://agentstack.voostack.com/s/hybridlabor-api
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.