Creator Platform Evidence Schema Interview

Assumptions Manifest

[from spec] The user wants more creator-media skills beyond YouTube.
[from codebase] Current YouTube skills are evidence-first and write artifacts under research/youtube/, with raw metadata/transcripts stored before synthesis.
[from codebase] Existing reusable creator strategy skills include creator-positioning, content-programming, series-spec, product-led-media-map, and creator-metrics-review.
[inferred] The new work should expand the pack, not replace the YouTube workflow.

[inferred] Create a spec for multi-platform creator-media audit skills.
[inferred] Prioritize skills that can run from durable evidence: official APIs, owner exports, RSS feeds, or explicitly user-provided analytics exports.

[from codebase] Skills are Markdown contracts mirrored under packs/creator-media/{codex,claude}/.../SKILL.md.
[from codebase] Existing audit style favors local files, explicit evidence paths, no fabricated transcript/metric claims, and archive-first replacement for canonical research docs.
[inferred] Avoid adding package dependencies unless a later implementation step proves a CLI/library is needed.

[from codebase] Changes would coexist with existing YouTube skills and likely add new platform-specific evidence collectors plus a cross-platform portfolio/metrics layer.
[inferred] Risk: if the pack overpromises YouTube-level metrics for platforms with limited APIs, the skills become unreliable or drift into scraping/ToS gray areas.

[inferred] Store raw platform evidence under research/<platform>/data/<slug>/ or a shared creator-platforms evidence directory.
[inferred] Use common normalized fields where available: post ID, URL, title/text, publish date, media type, duration/word count, views/impressions where available, likes/reactions, comments/replies, reposts/shares, saves/bookmarks, clicks, transcript/body text, thumbnails/media references.
[inferred] Missing metrics should be first-class evidence gaps, not silently omitted.

[from research] TikTok, Instagram, Threads, X, LinkedIn, beehiiv, podcast, Bluesky, Mastodon, and newsletter ecosystems have very different free/manual/API access surfaces.
[from research] LinkedIn personal export is viable for owner-provided personal evidence; LinkedIn organization analytics are admin/API-gated.
[inferred] The skill contract should distinguish public-data audits from owner-authorized audits.

[inferred] No credential capture in specs or research artifacts; use env vars or user-provided exports.
[inferred] No scraping-first workflows for platforms with fragile/closed access.
[inferred] Rate limits, missing fields, private metrics, and API eligibility should be explicit report sections.

Should the first implementation target be skills for specific platforms first, or a platform capability matrix plus shared evidence schema first?
Which platforms matter most for the creator pack?
Are owner-authorized/manual export workflows acceptable where official APIs are limited?
Can agents do web scraping?
Should the work use only free/open-source/manual ways to get information?

The user approved the recommended matrix-first approach.
The first implementation target should be a platform capability matrix plus shared evidence schema.
LinkedIn is the first named priority platform.
The user proposed a skill that tracks a content creator's career and presence as a whole in Markdown documentation held in a git repo.
Web scraping is acceptable to consider, but the final direction is free, open-source, and manual information collection.

Pros: faster visible pack expansion for LinkedIn or another platform.

Cons: risks duplicating incompatible schemas and locking in platform-specific assumptions too early.

Pros: makes platform capabilities explicit, avoids overpromising, and gives future skills a shared contract.

Cons: delays platform-specific audit implementation.

Recommendation: matrix and schema first. The user accepted this.

Pros: repeatable and structured where access exists.

Cons: platform approval, auth, rate limits, paid tiers, and missing personal-account access can block usage.

Pros: works without paid dependencies, fits repo-based documentation, and keeps the pack usable immediately.

Cons: lower automation and weaker metrics coverage.

Recommendation: free/manual/OSS baseline, with APIs only as optional future lanes. The user accepted this.

LinkedIn account export supports owner-provided personal data downloads, making it a viable baseline for personal creator/career evidence.
LinkedIn organization share statistics are available through official admin/API lanes, but this should be a later authorized workflow rather than a baseline.
trafilatura is an open-source Python/CLI tool for extracting readable text and metadata from web pages and saved HTML.
Playwright can capture full-page screenshots and HTML for public pages where visual evidence matters.
Open-source tooling should be optional and never used to bypass access controls.

Implementation goal covered: matrix/schema first, LinkedIn as first use case, creator presence dossier as strategic artifact.
Architecture covered: new skills feeding existing creator-media strategy skills.
Data model covered: shared evidence schema with capture methods, auth context, metrics object, and confidence fields.
API/contracts covered: output paths, evidence directories, collection methods, and skill names.
Edge cases covered: sensitive exports, blocked pages, missing metrics, missing body text, stale snapshots, identity aliases.
Security/privacy covered: no private scraping, no credential capture, redact sensitive exports.
Test strategy covered: metadata scans, targeted grep, schema fixture validation, diff checks.

The initial draft considered multiple platform-specific audit skills. The final spec prioritizes a capability matrix and schema foundation before platform skills.
API integrations were demoted from a possible baseline to optional future lanes.
LinkedIn is treated as an owner-export/manual evidence lane first because personal LinkedIn scraping and API access are constrained.
A new repo-based creator-presence-dossier artifact was added to track the creator's career and public/professional presence over time.