Skip to content
Agentmart
Submit a tool

td

Generate a Technology Disclosure (TD) document for a project.

Skill by cslev

IndexedMIT
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/cslev-td

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Generate a Technology Disclosure (TD) document for a project. A TD is a short (typically under 20 pages) research-paper-style document describing a software contribution, tool, framework, or technique developed within a project. It includes abstract, introduction, literature review, technical description, evaluation, and conclusion. Trigger when the user mentions "TD", "technology disclosure", "tech disclosure", "write up the disclosure", "patent disclosure", "invention disclosure", or asks for a research-paper-style writeup of a project, a component of it, or a specific technique. Also trigger when the user asks to review or critique an existing TD draft. Do NOT trigger for release documentation (the documentation skill handles that), pitch decks (the pitch-deck skill handles that), or general code documentation.

Use cases: documents, presentations, frontend, writing

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found
Runs as
Python package

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/td

What's inside

Skill name
td
Description
Generate a Technology Disclosure (TD) document for a project. A TD is a short (typically under 20 pages) research-paper-style document describing a software contribution, tool, framework, or technique developed within a project. It includes abstract, introduction, literature review, technical description, evaluation, and conclusion. Trigger when the user mentions "TD", "technology disclosure", "tech disclosure", "write up the disclosure", "patent disclosure", "invention disclosure", or asks for a research-paper-style writeup of a project, a component of it, or a specific technique. Also trigger when the user asks to review or critique an existing TD draft. Do NOT trigger for release documentation (the documentation skill handles that), pitch decks (the pitch-deck skill handles that), or general code documentation.
License
MIT
Instructions
304 lines in SKILL.md · view on GitHub

Before installing

  • Ships executable scripts

Files (5)

  • SKILL.md
  • build_td_docx.py
  • docx-generation.md
  • review-rubric.md
  • README.md · samples/

SKILL.md (first lines)

# Technology Disclosure (TD) Skill

## On activation

**Always announce that this skill has been triggered.** At the very start of your first reply, output exactly one line:

> **TD skill activated.** I'll run a pre-flight, draft the disclosure in markdown, then step you through review, humanizing, and `.docx` generation — pausing for your go-ahead at each step.

Then proceed immediately to Phase 1. No further preamble.

---

This skill generates Technology Disclosures: short research-paper-style documents describing a software contribution from a project. The working format is markdown (for fast iteration). Generation of the final `.docx` is the **last** step and happens only after the user has approved the draft — never automatically right after drafting.

The flow is gated: draft → (ask) self-review → revise until satisfied → (ask) humanize → (ask) generate `.docx`. Each arrow is a checkpoint where you stop and wait for the user. Do not run ahead to the next step on your own.

A TD is **not**: release documentation, a pitch deck, a marketing piece, a code walkthrough. It is closer to an academic research paper but produced for internal IP/disclosure purposes — typically under 20 pages, with the structure and rigor of a short conference paper.

## Deliverables

Files written to the TD output directory, in the order they appear:

1. **Draft TD** — `<PROJECT>_TD_<short-title-slug>_<YYYY-MM-DD>.md` (always)
2. **Self-review** — `<PROJECT>_TD_<short-title-slug>_<YYYY-MM-DD>_review.md` (only if the user opts into review)
3. **Final TD** — `<PROJECT>_TD_<short-title-slug>_<YYYY-MM-DD>.docx` (only when the user approves generation, as the last step)

Where `<PROJECT>` is the project name from `../project_context.md`.

The markdown draft is the working artifact; it is revised in place across review and humanizing. The `.docx` is generated once, at the end, from the approved draft.

---

## Prerequisites

Both are needed for the `.docx` generation step (Phase 5). They do not need to be installed up front — that phase checks for them and offers to install what's missing (see `./docx-generation.md`):

- **`pandoc`** (`brew install pandoc` / `apt install pandoc`) — converts the markdown body and builds the Word table-of-contents field.
- **`python-docx`** (`pip install python-docx`) — adds the cover page, sections, header/footer, page numbering, and justification.

---

## Required reading before doing anything

Before writing any code, generating any draft, or running the review pass:

1. Read `../project_context.md` — authoritative facts about the project.
2. Run the Bootstrap check (see below) — creates output and samples directories if missing.
3. Check `docs/td/samples/` — see "Samples handling" below for what to do based on what you find.

If `../project_context.md` has placeholder `<fill in: ...>` markers still unfilled and they're relevant to what the TD is about, stop and tell the user — the TD cannot be reliable until those are real.

---

## Bootstrap — first-time setup check

Run this before anything else, every invocation:

1. Check whether `docs/td/` exists at the project root.
2. **If it does not exist**, create it along with the shared assets directory:
   ```bash
   mkdir -p docs/td/samples/ docs/assets/logos/
   ```
   `docs/assets/logos/` is a **shared** project asset directory (not specific to this skill — other skills, e.g. documentation, can use the same logos). Then tell the user:
   > "This skill requires `docs/td/` in your project root — created it now along with `docs/td/samples/` and a shared `docs/assets/logos/`. Two things to drop in:
   > - **Sample TDs** (`.pdf` / `.docx`) into `docs/td/samples/` — the skill calibrates its tone, structure, and citation style against them.
   > - **Your organization's logo** (`.png` / `.jpg`) into `docs/assets/logos/` — it goes on the `.docx` cover page and in the page header.
   >
   > Re-invoke me once you've added them, or tell me to proceed now (I'll work without samples — output may not match your house style — and ask about the logo when it's time to generate the `.docx`)."

   Wait for the user's response before continuing to Phase 1.

3. **If `docs/td/` exists** but `docs/td/samples/` or `docs/assets/logos/` do not, create the missing ones silently and proceed.

---

## Samples handling

The `docs/td/samples/` directory holds reference TDs from the user's organization. They are used by both the drafter (for tone, structure, conventions) and the reviewer (for style reference during the self-review pass).

**If `docs/td/samples/` contains usable files** (`.pdf`, `.docx`, or `.md` files; ignore `README.md`): read all of them carefully. They are the calibration target for tone, section weight, citation style, table use, and figure conventions.

**If `docs/td/samples/` is empty**: ask the user before proceeding:

> "I don't see any sample TDs in `docs/td/samples/`. The skill works best with at least one reference TD from your organization so the output matches your house style. You can:
> (a) drop one or more `.pdf` / `.docx` TDs into `docs/td/samples/` and re-invoke me, or
> (b) tell me to proceed anyway and I'll produce a best-effort TD using abstract conventions only (output may not match your org's specific style).
>
> What would you like to do?"

If the user picks (b), proceed but include a brief note in the hand-off summary mentioning that the TD was generated without samples and may not match house style. Do not silently lower quality without flagging.

**If `docs/td/samples/` contains a `README.md`**: read it — it may contain user-supplied notes about which samples illustrate which patterns, or about house-convention deviations from the samples (e.g. "samples name commercial competitors but house convention is generic descriptions"). These notes override what you'd infer from the samples alone.

---

## Output directory

Always use `docs/td/` for output and `docs/td/samples/` for reference samples. These paths are fixed — do not look for or create alternative locations.

---

## Workflow

The skill has six phases, each gated on the user's go-ahead:

1. **Pre-flight** — gather scoping context.
2. **Draft** — write the markdown draft. Then *ask* whether to run the self-review.
3. **Self-review** — if the user agrees, review and then revise the draft until they're satisfied.
4. **Humanize** — *ask* whether to apply the `/humanizer` skill to the approved draft.
5. **Generate `.docx`** — *ask* whether the user is ready; then handle the logo and build the document.
6. **Hand off** — summarize and point to the files.

Do them in order. Do not skip a checkpoint or run the next phase before the user says to.

---

### Phase 1 — Pre-flight (ask the user)

TDs are scoped, audience-aware documents. Ask the user the following before drafting. This is "ask more" mode — gather enough context that the draft is grounded, not generic. Bundle the questions; don't ask one at a time.

From cslev/skillet, MIT.

Signals

Stars
0
Forks
0
Open issues
0
Last commit (this folder)
Jun 26, 2026 (4 months ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from cslev

Cite this page

td skill. Agent Mart. https://agentmart.dev/skills/cslev-td (last updated 2026-10-24)