franklee16

Fact Refresh — Security skill for Claude Code

Security community

Claude Code skill: audit course materials (Beamer decks, lecture notes, syllabi) for outdated facts, statistics, dates, rankings, and regulations; research current values from primary sources; apply s.

How to install Fact Refresh

This entry records only its repository, not the path inside it, so there is no exact command to give. Open franklee16/fact-refresh and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Fact Refresh does

Claude Code skill: audit course materials (Beamer decks, lecture notes, syllabi) for outdated facts, statistics, dates, rankings, and regulations; research current values from primary sources; apply surgical updates with compile verification

Alternatives in Security

  • Review Artifact — Creates one polished, self-contained HTML artifact from review findings, audit notes, PR feedback, code review 652 ★
  • Engage.Cvediff — Find the canonical fix commit(s) for a CVE across sources, then diff for root cause 348 ★
  • Beamer Skill — A Claude Code skill for creating, compiling, reviewing, and polishing academic Beamer LaTeX presentations 163 ★

README

fact-refresh

Audit a folder of course materials (Beamer decks, lecture notes, syllabi, handouts) for outdated facts, statistics, dates, rankings, and regulations — then research current values from primary sources and apply the updates with compile verification.

A two-stage workflow: **audit first, change nothing** → user reviews the punch list → **research and apply**, with surgical edits and source arbitration. Designed for term-prep runs 1–3 months before a new offering.


What it does

The skill runs in two clearly separated stages with a mandatory approval gate between them:

**Stage 1 — Audit (read-only)**

  • Dispatch one scan agent per unit (week / chapter / lecture), in parallel.
  • Each agent reads its unit and emits findings across a fixed 12-category taxonomy with MUST / SHOULD / NICE severity.
  • A consolidation agent produces a master punch list (final_update_list.md) with universal fixes first, then per-unit items, with effort estimates.
  • Stops. Presents the punch list. Waits for explicit user approval before editing anything.

**Stage 2 — Research + Apply (after approval)**

  • Phase A — Research. Dispatch domain-clustered research agents (e.g., rates/bonds; equities/funds; FX; derivatives; regulation) that pull current values from primary sources. Each writes its own data file.
  • Phase B — Apply. Dispatch per-unit edit agents that read their findings + the relevant research file and edit source files surgically, with coordinated recompute of chained worked examples and compile verification.
  • Phase C — Verify. Confirm fresh PDFs, write a fix_summary.md with universal fixes, wrong-direction corrections (wrong, not just stale), deferred items, and a professor-arbitration list for source conflicts.

When to trigger

Use any of:

  • Natural language: "check my slides for outdated info", "refresh the data in my course materials", "update the facts in my decks before the term", *"what needs updating in my