skill-market-publisher

1---
2name: skill-market-publisher
3description: Publish, submit, and verify local skills across public skill marketplaces, directories, and registries. Use when an agent needs to take a skill folder or skill repository, confirm current submission paths, prepare cross-market metadata, run verified live adapters, or generate manual submission bundles for markets such as AgentSkill.sh, Skillstore, SkillNet, agent-skills.md, SkillsMD, SkillHub, SkillHQ, and related registries.
4---
5 
6# Skill Market Publisher
7 
8## Overview
9 
10Use this skill to turn one local skill folder into a reusable market release plan. Keep the skill itself timeless: document current submission surfaces, verification patterns, and adapter logic, but keep one-off submission IDs, sample URLs, and dated field logs outside the reusable skill folder.
11 
12## Quick Start
13 
141. Re-check live market surfaces before making claims about support:
15 
16```bash
17python3 scripts/skill_market_publish.py recon
18```
19 
202. Inspect one skill and build a market readiness plan:
21 
22```bash
23python3 scripts/skill_market_publish.py plan /abs/path/to/skill \
24  --repo-url https://github.com/owner/repo \
25  --git-ref main \
26  --tag automation \
27  --tag research
28```
29 
303. Generate a review bundle before any live submission:
31 
32```bash
33python3 scripts/skill_market_publish.py bundle /abs/path/to/skill \
34  --repo-url https://github.com/owner/repo \
35  --git-ref main \
36  --author-name "Your Name" \
37  --author-email you@example.com \
38  --version 0.1.0 \
39  --skillz-category automation \
40  --bogen-category development \
41  --a2a-category research \
42  --a2a-price 9.0 \
43  --a2a-seller 0xYourWalletAddress \
44  --out-dir /tmp/skill-market-bundle
45```
46 
474. Publish one wrapped live target at a time. Live submission requires `--execute`.
48 
49```bash
50python3 scripts/skill_market_publish.py publish agentskill-sh /abs/path/to/skill \
51  --repo-url https://github.com/owner/repo \
52  --execute
53```
54 
555. Verify the result using the market-specific checks in [references/verification-playbook.md](references/verification-playbook.md).
56 
576. When a market still needs browser-side login or form interaction, open the tracked submit pages and use the printed checklist:
58 
59```bash
60python3 scripts/open_manual_submit_pages.py \
61  --repo-url https://github.com/owner/repo \
62  --git-ref main \
63  --skill-path /abs/path/to/skill-a \
64  --skill-path /abs/path/to/skill-b
65```
66 
67## Workflow
68 
69### 1. Normalize the Release Unit
70 
71Treat the publishable unit as one skill folder containing `SKILL.md`.
72 
73- If the repository contains many skills, publish each skill separately.
74- If the repository is a pack, split markets into:
75  - pack importers that should ingest the public repo once
76  - per-skill submitters that should receive a folder URL or raw `SKILL.md` content
77- If a market works at repository scope, make sure the chosen public repository layout matches that market before submitting.
78 
79### 2. Detect Repository Layout
80 
81Run `inspect` or `plan` first. The bundled CLI classifies the current repository as:
82 
83- `single-skill-repo`
84- `root-pack`
85- `skills-subdir-pack`
86- `nested-skill-dir`
87- `standalone-skill-dir`
88 
89Some markets accept any repository URL. Others currently prefer a repository with a top-level `skills/` directory. When layout and market expectations conflict, create a mirror or release repository instead of reshaping the source repository for one listing.
90 
91Treat submit-page copy as a hint, not proof. Some live backends accept `root-pack` repositories even when the visible page copy talks about a repo-root `skills/` directory. Verify layout assumptions from the live preview, submit response, or public read API before forcing a mirror.
92 
93### 3. Choose the Market Mode
94 
95Use automated publishing only for targets that are currently verified in [references/market-matrix.md](references/market-matrix.md).
96 
97Current modes:
98 
99- `auto-cli`: official CLI publishing wrapped by this skill
100- `auto-http`: stable HTTP form or API submission wrapped by this skill
101- `manual-web`: browser submission workflow with no stable documented endpoint in this skill
102- `index-only`: discovery depends on installability, GitHub App ingestion, or passive indexing rather than explicit submission
103- `incompatible`: expects a different artifact format than a plain skill folder
104- `needs-recon`: public site exists, but the current publish path is not verified
105 
106Some markets also expose official CLI or seller dashboards that this skill records but does not proxy. Treat those as documented manual flows unless the bundled CLI explicitly supports them.
107 
108### 4. Prepare Market Inputs
109 
110Provide explicit values for market-only fields instead of guessing.
111 
112- Use `--version` for ClawHub publishing.
113- Use `--author-email` and `--skillz-category` for Skillz Directory.
114- Map domain-specific categories onto the market's real enum. If the market does not expose `research`, `browser`, or another domain label you want, choose the closest supported value or `other` instead of inventing one.
115- Use `--a2a-price`, `--a2a-category`, and `--a2a-seller` for A2A Market.
116- Use `--git-ref` when the bundle should include stable GitHub `blob` and raw `SKILL.md` URLs, or when a market benefits from a folder-scoped GitHub `tree` URL.
117- When a market accepts a GitHub URL, decide deliberately between:
118  - repository URL
119  - repo-root pack URL
120  - folder `tree/<ref>/<path>` URL
121 
122Accepted input alone does not prove scoping. Some markets accept a folder URL but still ingest the entire repository pack.
123 
124Run `plan` after setting these fields. The CLI reports missing inputs and layout blockers per market.
125 
126### 5. Build A Review Bundle
127 
128Use `bundle` before live submission. The bundle gives you:
129 
130- normalized skill metadata
131- a per-market readiness matrix
132- payloads for wrapped live targets
133- manual submission notes for every non-wrapped target
134 
135Use the bundle as an operator review packet, not just a payload dump. Before live submit, check that it exposes:
136 
137- the canonical repository URL and `SKILL.md` URL
138- the exact URL scope you intend to submit
139- one short example task or prompt for the listing
140- the strongest verification URLs you expect to use after submit
141 
142### 6. Publish Safely
143 
144Use this order:
145 
1461. `inspect`
1472. `plan`
1483. `bundle`
1494. `publish <market>` without `--execute`
1505. `publish <market>` with `--execute`
1516. verify with the market-specific public checks
152 
153Do not claim one-click support for a market marked `manual-web`, `index-only`, `incompatible`, or `needs-recon`.
154 
155Interpret live responses by semantics, not by HTTP status alone:
156 
157- `alreadyExists` or `updated` often means the market recognized an existing public listing.
158- `processing` or `submission received` means the intake pipeline accepted the request, not that the listing is public yet.
159- duplicate-name or pending-review errors can still prove the market recognized the listing identity.
160- a redirect to a public detail page is stronger evidence than a flash message or success banner.
161 
162### 7. Record Outcomes Outside The Skill
163 
164Keep reusable skill docs free of sample-specific evidence.
165 
166- Do not store live submission IDs, PR numbers, sample slugs, or dated field logs inside this skill.
167- Put one-off execution evidence in the task conversation, a separate operator notebook, or a temporary bundle outside the reusable skill folder.
168- Update this skill only when the reusable workflow changes.
169 
170## Bundled CLI
171 
172### `inspect`
173 
174Parse frontmatter, derive summary text, detect repository layout, and show normalized metadata.
175 
176### `plan`
177 
178Build a readiness matrix for known markets and report missing fields or layout warnings.
179 
180### `bundle`
181 
182Write a reusable submission bundle containing:
183 
184- `manifest.json`
185- `market-plan.json`
186- per-market payload files
187- manual submission notes
188 
189### `publish <market>`
190 
191Supports live execution only when `--execute` is present.
192 
193Wrapped live targets:
194 
195- `clawhub`
196- `agent-skills-md`
197- `agentskill-sh`
198- `skillz-directory`
199- `skillstore-io`
200- `skills-re`
201- `skillsmd-dev`
202- `bogen-ai`
203- `skillsrep`
204- `a2a-market`
205 
206### `recon`
207 
208Fetch known market pages and check expected markers. Run this when a marketplace may have changed since the last verification pass.
209 
210### `open_manual_submit_pages.py`
211 
212Open the current manual-web submit pages in the default browser, print a short per-site checklist, and optionally print exact folder and `SKILL.md` URLs for the skills being submitted.
213 
214## Operating Rules
215 
216- Publish from a public repository URL unless the market is local-CLI only.
217- Keep skill content English-only unless a target market explicitly requires localized marketing copy.
218- Prefer stable repository URLs and stable `SKILL.md` URLs when the market supports deep links.
219- When a market expects a different repository layout, create a publish mirror instead of mutating the source repository just for one listing.
220- Prefer the strongest public verification surface the market exposes: status API, workflow URL, detail page, public read API, then search page.
221- If a folder URL submission returns a pack-sized import count, treat the scope as repository-wide until the public result proves otherwise.
222- Keep the reusable skill free of sample submission logs, IDs, and dated evidence.
223- Treat market behavior as time-sensitive. Re-run `recon` before changing adapters or writing claims about current support.
224 
225## References
226 
227- Read [references/market-matrix.md](references/market-matrix.md) when choosing markets or checking current compatibility.
228- Read [references/publish-playbook.md](references/publish-playbook.md) when mapping a skill repo to each market's intake model.
229- Read [references/verification-playbook.md](references/verification-playbook.md) when checking whether a submission is public, pending, blocked, or only indexed.
230- Read [references/recon-playbook.md](references/recon-playbook.md) when adding a new market or repairing a broken adapter.
231