Skip to content

v4.0.3 — SKILL.md description compliance (Issue #9)

Latest
Latest

Choose a tag to compare

View all tags
@pitimon pitimon released this 11 May 21:57
v4.0.3
6a1df52
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

What's Fixed

Closes #9 — skills/cybersecurity-pro/SKILL.md description: frontmatter trimmed from ~4,663 → 590 characters.

Compliance

Cap Limit v4.0.2 (#9 reported) v4.0.3
Anthropic SKILL.md spec 1024 4,663 ✗ 590 ✓
Claude Code skillListingMaxDescChars (default) 1536 4,663 ✗ 590 ✓
writing-skills recommended ≤500 4,663 ✗ 590 (≈recommended)

Impact

  • /doctor warning resolved — no more "descriptions exceed the per-entry cap" notice on every session
  • No more silent truncation of trailing trigger keywords (previously Claude Code clipped at the cap boundary with no signal which keywords were lost)
  • ~15k tokens/session overhead eliminated for users who opted in to skillListingMaxDescChars: 4096 to retain the full string
  • Anthropic spec compliance restored — risk that Claude follows the summary instead of reading the full skill body is gone

What moved where

The long trigger keyword list (15 categories spanning Incident Response, DevSecOps, Cloud, AI/ML, OT/ICS, PQC, Identity, Web3, Thai prompts, and Shannon handoff) was moved from the frontmatter description: to the skill body under a new section "When This Skill Activates / เมื่อใดที่สกิลนี้ทำงาน". Claude reads the body when the skill is invoked anyway, so trigger coverage is preserved without consuming session-listing token budget.

New "Use when…" description

Use when asked to generate professional cybersecurity documents — IR
playbooks, DFIR forensic reports, SOC L1–L3 triage runbooks, DevSecOps
pipeline configs, threat models, compliance gap analyses, or executive
cyber-risk reports. Covers 22 domains across AppSec / Cloud / OT / AI-ML /
API / Identity / Web3, bilingual Thai+English output mapped to NIST 800-53,
MITRE ATT&CK, OWASP, ISO 27001. Also: post-pentest defensive (Shannon
handoff manifest). Triggers: "incident response", "SOC triage", "DFIR",
"threat hunt", "compliance audit", "การตอบสนองต่อเหตุการณ์",
"post-pentest defensive".

Other Changes

  • README.md v4.0.3 rework — Table of Contents, NIST CSF 2.0 coverage map, Frameworks (73) badge, structured Support table, Related Plugins refresh
  • Manifest version reconciliation — plugin.json / marketplace.json bumped to 4.0.3 (had remained at 4.0.1 when v4.0.2 tag was published on 2026-03-01)

Verification

  • bash tests/validate-plugin.sh --skip-install-check — 67 PASS / 0 FAIL (2 advisory framework-staleness WARN: mitre-attack, cisa-kev — unrelated to this release)
  • Folded YAML description length verified at 590 chars
  • All 22 domain references and trigger keyword categories preserved in skill body

PRs

  • #10 — fix: trim SKILL.md description to 590 chars + README v4.0.2 rework
  • #11 — chore: bump version to 4.0.3, reconcile manifest with tag history

Credit

Filed by a user hitting the /doctor warning during a normal session — thank you for the detailed reproduction steps and the impact analysis. 🙏

Assets 2
Loading