Claude Fable 5 Skills — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Claude Fable 5 Skills (Plugin) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
10 battle-ready Agent Skills for Claude Fable 5 and 5.1 — the first skill collection designed around how Claude Fable 5 actually behaves, not how older models needed to be babysat.
Anthropic's own guidance for Fable 5 notes that skills written for prior models are often too prescriptive for it and can degrade output quality. Old skills micromanage; Fable 5 needs intent, boundaries, and verification hooks instead. Every skill here is written Fable 5-first.
Works with Claude Code, Claude Cowork, Cursor, Copilot, Gemini CLI, and any harness that reads SKILL.md files (agentskills.io-style: YAML frontmatter + Markdown body).
| # | Skill | What it fixes |
|---|---|---|
| 1 | skill-refactorer | The meta-skill. Audits your pre-Fable-5 skills/prompts (and Fable 5 skills now running on 5.1), deletes capability-compensation scaffolding, keeps real guardrails. |
| 2 | act-when-ready | Over-planning at high effort: re-deriving settled facts, surveying options it won't pursue. |
| 3 | effort-calibrator | Picking the right effort level per workload; Fable 5 at lower effort often beats older models at xhigh, and 5.1 needs its own sweep plus per-message effort changes. |
| 4 | no-gold-plating | Diffs bigger than the ask: unrequested refactors, speculative abstractions, impossible-state error handling, and (on 5.1) test files out of proportion to the change — Anthropic reports the scoping instruction cuts extras substantially with no change in task success. |
| 5 | grounded-progress | Status reports on long runs must point at tool-result evidence — no more "tests passing" that never ran. Includes the harness setting that makes 5.1 progress updates visible. |
| 6 | scope-guard | "Diagnose" ≠ "fix". No unrequested actions, no state changes on pattern-matched evidence. |
| 7 | subagent-orchestration | Parallel delegation, long-lived workers, fresh-context verifier subagents that out-perform self-critique, and the per-turn nudge that keeps 5.1 batching tool calls. |
| 8 | markdown-memory | A file-based lesson memory Fable 5 exploits unusually well — with the maintenance discipline that keeps it useful. |
| 9 | autonomous-continuation | Unattended runs that stall on "I'll now run X" or mid-run permission questions, or quietly shrink a partly blocked task. Includes the context-budget composure pattern. |
| 10 | regrounding-summary | Final reports readable by someone who saw none of the work — no arrow chains, no invented abbreviations, no unmarked quotations, no dense unbroken paragraphs. |
Claude Code (recommended) — install as a plugin, get updates automatically:
/plugin marketplace add kpab/claude-fable-5-skills
/plugin install fable5-skills@claude-fable-5-skillsManual — copy any skill folder into your skills directory:
git clone https://github.com/kpab/claude-fable-5-skills
cp -r claude-fable-5-skills/skills/act-when-ready ~/.claude/skills/Or cherry-pick per project by placing folders under .claude/skills/ in the repo. Other harnesses: point your skill loader at the skills/ directory; each skill is a self-contained folder with one SKILL.md.
Four skills are task-shaped and trigger automatically when a matching request appears: skill-refactorer, effort-calibrator, subagent-orchestration, markdown-memory.
The other six are standing behavioral rules. A model that is mid-over-planning won't reliably decide to invoke act-when-ready on its own, so don't count on auto-triggering for these. Two reliable paths:
/act-when-ready) at the start of the session, or the moment the symptom shows up.CLAUDE.md; this is how Anthropic's own Fable 5 guide ships these patterns. autonomous-continuation is written for exactly this.Fable 5 shipped on 2026-06-09; Fable 5.1 shipped on 2026-09-01 and is now the current Fable, with Fable 5 listed as a legacy model. Anthropic states that existing Fable 5 prompts perform well on 5.1 without changes, and documents a handful of behavior differences; these skills fold those in (effort sweeps, test bloat, fewer progress updates, denser prose, one-call-per-turn loops). Expect further refinements as the community learns the model. Issues and PRs with reproducible before/after examples are very welcome.
This is an unofficial community project, not affiliated with or endorsed by Anthropic. Verify current API parameters against the official docs before production use. Fable 5.1 also carries API-level breaking changes outside the scope of these skills — forced tool_choice (any / tool) returns a 400, earlier models cannot read 5.1 thinking blocks, and editing earlier turns invalidates later thinking blocks — see What's new in Claude Fable 5.1.
MIT — see LICENSE.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.