agentleFS
grcmanagrcmana/grcskills/skills/policy-gap-review/SKILL.md

policy-gap-review

Skill2 starsChanged 15 days agoMITView on GitHub

Review existing policies against ISO 27001, SOC 2, Cyber Essentials, GDPR and ISO 42001 requirements and auditor expectations — flagging missing coverage, untestable statements, unkeepable commitments, missing document control, and contradictions with actual practice. Use for "review our policies", "gap check this policy", "are our policies audit-ready", "policy review", or annual policy review cycles.

3 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read references/grc-workspace.md and references/writing-standards.md first.

    line 17
  2. Distinctive:

    B. Document control: Check for:

    line 32
  3. Distinctive:

    Don't rewrite policies wholesale unless asked.

    line 72

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Policy Gap Review

You review policies the way a certification body auditor or SOC 2 auditor would: coverage first, then quality, then consistency with reality.

Read references/grc-workspace.md and references/writing-standards.md first.

Inputs

  • Policies in grc/policies/, or files or folders the user points to (Markdown, DOCX, PDF). Read them in full.
  • grc/context/company-profile.md, for the framework selection and actual practices.
  • Optional: an evidence tracker or risk register, to check that policy claims match reality.

Modes

  • set: review the whole policy set for coverage against the selected frameworks. This is the default.
  • single: deep review of one policy.
  • annual: the annual review cycle. Check currency, changes in the organisation or frameworks, and review and approval dates.

Review checks

A. Coverage (set mode). Map each policy to the domains in references/crosswalk.md and the expectations in references/policy-catalogue.md. Report uncovered requirements per framework, e.g. "No supplier security coverage → ISO 27001 A.5.19–A.5.23, SOC 2 CC9.2".

B. Document control. Check for:

  • an owner
  • an approver and approval date
  • a version
  • a review cycle, and that it's not overdue
  • a classification
  • revision history

C. Statement quality. For each statement, check:

  • Testability: can an auditor ask "show me"?
  • Keepability: frequencies and SLAs that look unrealistic for this organisation's size. Flag these as Type II exception risks.
  • Specificity: vague words ("appropriate", "regularly", "where possible") with no definition
  • Policy vs procedure: procedural detail that should move out of the policy

D. Consistency.

  • Contradictions between policies (e.g. password rules that differ).
  • Contradictions with practice in the company profile, e.g. the policy says "MDM on all devices" but the profile says BYOD without MDM.
  • Under Cyber Essentials, any policy permitting something the scheme fails on (unsupported software, no MFA on cloud services, patching slower than 14 days).

E. Framework-specific must-haves. For example:

  • ISO 27001 5.2 policy content: objectives framework, commitment to requirements and continual improvement
  • GDPR: 72-hour breach notification, and data subject request handling
  • ISO 42001: AI policy aligned to the organisation's AI role

Severity

  • Major: a missing required coverage, or a contradiction with practice that would likely cause a nonconformity or exception.
  • Minor: weak wording, missing document control, an overdue review.
  • Observation: an improvement opportunity.

Outputs

  1. grc/policies/gap-review-YYYY-MM-DD.md with:
    • an executive summary (3–5 lines)
    • a coverage matrix: domain × framework → policy or gap
    • a findings table: id, policy, section, finding, severity, framework_refs, recommendation
    • quick wins: fixable in under an hour each
  2. Offer to fix specific findings using the policy-writer conventions. Make edits only with the user's go-ahead, and bump the version and revision history.
  3. A run log in grc/runs/.

Guardrails

  • Review against framework expectations, and never claim a policy "complies".
  • Don't rewrite policies wholesale unless asked. Propose targeted fixes.
  • Quote at most short fragments of the user's own policies, which is fine. Never quote standard text.

References

  • references/grc-workspace.md
  • references/writing-standards.md
  • references/policy-catalogue.md
  • references/crosswalk.md
  • Framework packs: references/iso27001.md, soc2.md, cyber-essentials.md, gdpr.md, iso42001.md

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.