prysai/prysai-llm-playbook/prysai-platform-fact-watch
Claude Code and agent skill by prysai/prysai-llm-playbook
- community
Checking that the SKILL.md file is still in the repository.
Maintain current, named-platform teaching claims after a vendor document, product surface, permission, model, account path, or link may have changed. Use when a curriculum maintainer needs to identify affected chapters, Labs, Skills, routes, and temporary reader-facing limits for Codex, Claude Code, Grok, ChatGPT, Gemini, Copilot, or another named LLM platform. Do not use to retrieve facts, admit a new adapter, execute a platform, or compare models.
Install
npx vskill@latest install prysai/prysai-llm-playbook/prysai-platform-fact-watchSKILL.md
Platform Fact Watch
Turn “this platform may have changed” into a small maintenance decision. This Skill inventories existing claims and their blast radius. It does not browse, run a product, make a release, or replace a source review.
Start with a claim card
Require one named platform, one source-backed claim or claim ID, its current reader-facing location, source owner and URL, last check date, scope, owner, next review, and the reason for review. Treat a missing field as unreviewed, not as a harmless blank.
Keep the claim narrow. “Claude Code has a permission mode” and “Grok Build has an API route” are separate cards. Do not use a platform name, a feature label, or an HTTP response as a substitute for a claim.
Classify the change signal
Choose one status without inferring current product behavior:
review_due: the scheduled review date has arrived or the source has not
been checked at the declared interval;
source_changed: a dated first-party source review reports a material
difference from the recorded claim;
source_unavailable: the cited source cannot presently support the claim;scope_changed: the claim may no longer apply to the named surface, account,
region, version, or permission boundary;
no_change_recorded: a dated first-party review found the same scoped claim;unreviewed: no suitable first-party review is available.
Do not choose no_change_recorded from memory, a redirected URL, a search snippet, a community post, or a successful sign-in. A source check confirms a statement only within its recorded date and scope.
Map the affected teaching surface
List each affected canonical unit and label its role:
Stable-core principles such as explicit authority, evidence, recovery, and least side effect normally remain usable. A product command, UI path, permission default, price, entitlement, integration, or model availability is an adapter fact and needs source review. Do not turn a source change into a claim that the whole course is broken.
Choose the smallest safe action
no_change_recorded: keep the scoped wording and update only the review
receipt; do not claim broader durability.
review_dueorunreviewed: retain the universal core, mark the named step
for review, and hand off the current fact to prysai-source-investigator.
source_changed,source_unavailable, orscope_changed: pause or remove
the named instructional step until a source review establishes replacement wording. Preserve the prior record as historical evidence.
- if the change calls the adapter’s source, run, authority, or failure record
into question: hand off the admission decision to prysai-platform-adapter-review.
- if a public claim, generated page, or release note already states the old
fact: hand off the artifact packet to prysai-evidence-review before publishing a correction.
Never silently rewrite a product procedure from memory. Do not classify an adapter as admitted, safe, equivalent, or production-ready from a freshness receipt.
Return a maintenance receipt
Return exactly one record containing the claim card, change status, affected units, reader risk, safe interim text, source-review handoff, any adapter or claim-audit handoff, owner, next review, and unknowns.
End with this limit: This receipt manages the freshness boundary of one named platform claim. It does not prove current product behavior, account access, permission safety, runtime success, adapter admission, model quality, learner outcome, or cross-platform equivalence.
Maintenance record
source: original Prysai Lab maintenance method derived from ADR-0025, the
content lifecycle, the fact-impact registry, and the source-bounded adapter admission record
license: original rewrite; first-party platform documentation and public
reports remain reference-only under docs/sources/asset-register.md
owner: facts-maintainerversion:0.1.0review_date:2026-09-14content_status:candidate
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/prysai-platform-fact-watch/SKILL.md
Popularity
- Repository stars
- 437
- Repository forks
- 8
- 7-day trend score
- 30
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/prysai/prysai-llm-playbook/prysai-platform-fact-watch)