360 Secure SkillHub
360 安全技能中心

audit

该 Skill 用于对照用户的权威职业事实与已批准的品牌指南,审计简历、LinkedIn、GitHub、作品集或网站等职业展示面,识别事实冲突、呈现问题和视觉漂移;单面输出发现报告,多面额外生成差异矩阵与去重后的行动项。它不会编辑或发布任何展示面,也不处理 SEO、无障碍、性能或安全审计。

分类:数据
lfarci
热度 0.00
最近更新 2026-10-30 05:10:27

安装

发给 Agent
帮我安装这个技能:npx secure-skills-cli add lfarci/career-skills/audit
兼容性
通用兼容
联网要求
不联网
数据出网
数据不出本机
Metadata
name
audit
description
Audit one or several professional surfaces (resume, LinkedIn, GitHub, portfolio, website) against career facts and brand guidance. Use for checking a profile's claims, consistency or visual drift.

Audit Professional Surfaces

Check what a surface actually shows against the user's canonical facts and approved brand guidance. One surface gets a findings report; several surfaces also get a cross-surface matrix that finds relationships, not concatenated reports. This is not an SEO, accessibility, performance or security audit.

Contract

  • Input: One or more surfaces from brand/surfaces.md or supplied directly (URL, export or pasted text), plus relevant profile/ facts, brand/positioning.md, brand/identity.md and staged assets.
  • Scope: Find factual, presentation and visual-drift issues; ask how to reconcile differences instead of assuming live or staged content is right. Never edit surfaces or overwrite records.
  • Output: Findings in the response or the user's existing review/tracker record; with several surfaces, a discrepancy matrix and deduplicated actions.

Steps

  1. Scope. Identify each surface, accessible version, date checked and audience. Mark inaccessible parts unknown; accept exports and pasted content. In an export archive (for example a LinkedIn data zip) read only the profile files the audit needs, in place; leave messages and connections unopened.
  2. Extract claims by subject: headline, roles and dates, experience, skills, credentials, projects, links, contact. When visible, note banner, avatar/logo, palette and typography.
  3. Compare with canonical records. Classify each difference as confirmed mismatch, needs verification, valid tailoring or stylistic opportunity. Compare text with the staged copy and the surface's Emphasise: line, and visuals with brand/identity.md when it exists; platform crops and font fallbacks are not drift.
  4. Several surfaces only: build an alignment matrix (canonical value, each surface's wording or omission, classification), add a visual-identity row when relevant, and check real links between surfaces (site → resume, GitHub → portfolio). Trace root causes and fix source facts before derived copies.
  5. Credential surfaces (Credly, Microsoft Learn transcript, similar): a certification or badge that is missing or outdated in profile/qualifications.md becomes a proposed profile update (name, issuer, date, evidence URL). Ask the user; apply only an approved change, following maintain-profile's record rules. A recurring check is the host's scheduling, not this skill.
  6. Report. Prioritize by risk. For each actionable item give surface(s), current state, evidence, desired outcome and a verifiable acceptance criterion. Check the tracker for existing issues before proposing new ones; write to a tracker only when authorized, otherwise output issue-ready Markdown.

Done when

  • Every accessible in-scope surface was inspected; inaccessible content is reported unknown, not assumed aligned.
  • Confirmed contradictions cite evidence; tailoring and platform variants are not flagged as errors.
  • Planned content is not treated as live.
  • Every high-priority finding has a measurable correction; no duplicate items.
  • Credentials found only on a surface are proposed as profile updates, never written silently.
  • Nothing was published or written to a tracker without authorization.