FazBrowse GitHub Viewer | Trending |
URL:
| Home
Tools: [Download Repo ZIP]   [Original HTTPS Page]

docs: front-load "YouTube MCP" / "YouTube Skill" across discovery surfaces by therohitdas · Pull Request #8 · ZeroPointRepo/transcriptapi-plugin · GitHub

docs: front-load "YouTube MCP" / "YouTube Skill" across discovery surfaces - #8

Merged
therohitdas merged 1 commit into
mainfrom
docs/frontload-keywords
Aug 23, 2026
Merged

docs: front-load "YouTube MCP" / "YouTube Skill" across discovery surfaces#8
therohitdas merged 1 commit into
mainfrom
docs/frontload-keywords

Conversation

Copy link
Copy Markdown
Contributor

Every place this package gets discovered led with framing instead of with the words people type. Someone searching "youtube mcp" matched none of them early.

Reordered

Surface Before After
GitHub About "YouTube for AI agents as one Agent Plugin — … 13 skills" "YouTube MCP + YouTube Skill for AI agents — …"
README h1 "TranscriptAPI — YouTube for AI Agents" "YouTube MCP + YouTube Skill for AI Agents"
Manifest description ×7 "YouTube transcripts, captions, …" "YouTube MCP + YouTube Skill for AI agents — …"
Codex displayName "TranscriptAPI — YouTube for AI Agents" "YouTube MCP + Skill — TranscriptAPI"
keywords youtube, transcript, … youtube-mcp, youtube-skill hoisted to 1 and 2
topics 15 +youtube-mcp, +youtube-skills

Brand isn't lost — it moved to the line directly beneath the h1.

The old About still advertised "13 skills", stale since the consolidation. That was the more damaging half: it's the single most-read sentence in the repo.

About and topics are repo metadata rather than files, so those were set via the API and aren't in this diff — they're already live.

What I deliberately did not touch

skills/youtube/SKILL.md's description. That field is the agent's trigger-matching surface, not marketing copy — it's the only thing loaded at startup when deciding whether to activate the skill. Stuffing "YouTube MCP" into it would bias activation toward the literal phrase and away from the natural-language phrasings it's tuned for. SEO and trigger vocabulary are different jobs and shouldn't share a field.

Also: fixed the defect I flagged

Step 3 asserted format="text" is the default, unqualified. True on MCP — but the REST default is json, and REST also defaults send_metadata to false.

references/rest-api.md already warned about this, but references load on demand, so an agent reading only SKILL.md and taking the REST path would omit the parameter and get JSON with no metadata. Now states both defaults explicitly.

Verification

ajv valid ×2 · skills-ref 1/1 · check_conformance.py 11/11 · all JSON/YAML parse · skill description unchanged at 732/1024 · SKILL.md 131/500 lines · all 8 manifests agree on 1.2.0.

🤖 Generated with Claude Code

…faces

Every place this package gets discovered led with framing instead of with the
words people type. The GitHub About opened "YouTube for AI agents as one Agent
Plugin", the README h1 opened with the brand, and the manifest description —
the text plugin browsers show in search results — opened the same way. Someone
searching "youtube mcp" matched none of them early.

Reordered so the searched terms come first, without inventing claims:

- README h1: "TranscriptAPI — YouTube for AI Agents"
  -> "YouTube MCP + YouTube Skill for AI Agents", with the brand moved to the
  line directly beneath so attribution is not lost
- manifest description in plugin.json and all four per-client manifests, plus
  both marketplace indexes
- .codex-plugin interface displayName and shortDescription
- keywords: youtube-mcp and youtube-skill hoisted to positions 1 and 2

The GitHub About and repo topics are repo metadata, not files, so those were
set through the API rather than in this commit. The old About still advertised
"13 skills" — stale since the consolidation, and the more damaging half of the
problem, since it was the single most-read sentence in the repo.

Deliberately did NOT touch skills/youtube/SKILL.md's `description`. That field
is the agent's trigger-matching surface, not marketing copy — it is the only
thing loaded at startup when deciding whether to activate the skill. Stuffing
"YouTube MCP" into it would bias activation toward the literal phrase and away
from the natural-language phrasings it is tuned for. SEO and trigger vocabulary
are different jobs and should not share a field.

Also fixed a real defect while in the file. Step 3 asserted `format="text"` is
the default, unqualified. That holds on the MCP path only — the REST default is
`json`, and REST also defaults `send_metadata` to `false`. references/rest-api.md
already warned about exactly this, but references load on demand, so an agent
reading only SKILL.md and taking the REST path would omit the parameter and get
JSON with no metadata. Now states both defaults and says to pass them explicitly
on REST.

Bumped to 1.2.0 — new discovery copy plus a behavioural correction in the skill.

Verified: ajv `valid` for both manifests; skills-ref 1/1; check_conformance.py
11/11; all JSON and YAML parse; skill description unchanged at 732/1024;
SKILL.md 131/500 lines; all eight manifests agree on 1.2.0.

Co-Authored-By: CRHQ <noreply@crhq.ai>
therohitdas merged commit c2df0d1 into main Aug 23, 2026
3 checks passed
therohitdas deleted the docs/frontload-keywords branch August 23, 2026 17:25
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants


Back | FazBrowse Home | New Git URL