fedramp-specialist
FedRAMP source of truth. Use whenever FedRAMP comes up in any form — FedRAMP 20x or Rev5, the Consolidated Rules for 2026, certification classes A–D, Program vs Agency paths, Key Security Indicators (KSIs), Marketplace listing, assessor or advisor obligations, agency sponsorship, continuous monitoring, significant change, vulnerability or incident timelines, FedRAMP public notices (NTC-), CISA BOD 26-04 for cloud providers, scope, OMB M-24-15, the FedRAMP Authorization Act — and also when someone asks how a SaaS or cloud product gets approved to sell to federal agencies, whether SOC 2 or GovRAMP counts toward FedRAMP, or what a "FedRAMP advisor" must do, even if the word FedRAMP is not used. Answers from the official FedRAMP 2026 markdown corpus and machine-readable rules bundled with this skill (auto-refreshed from GitHub and fedramp.gov), with FedRAMP notices and cited CISA/OMB documents as a labelled secondary tier, citing rule IDs and file paths, never from memory or third-party summaries.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/kamranarshad33--fedramp-specialist)HTML
<a href="/listings/kamranarshad33--fedramp-specialist"><img src="/api/listings/kamranarshad33--fedramp-specialist/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
Requires Claude Code, `git`, `bash`, and Python 3.9 or later (standard library only).
git clone https://github.com/kamranarshad33/fedramp-specialist.git ~/.claude/skills/fedramp-specialist bash ~/.claude/skills/fedramp-specialist/scripts/sync.sh --force
The repository does not include the three upstream FedRAMP repositories; the first sync clones them into `corpus/`. After that, the sync pulls the latest FedRAMP repositories, refetches the schemas and notices, rebuilds the indexes, and writes the subagent to `~/.claude/agents/fedramp-specialist.md` with paths for your machine. To write the subagent somewhere else, set `FEDRAMP_AGENT_PATH` first.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
SKILL.md · Manifest · 192 lines
- ---
- name: fedramp-specialist
- description: FedRAMP source of truth. Use whenever FedRAMP comes up in any form — FedRAMP 20x or Rev5, the Consolidated Rules for 2026, certification classes A–D, Program vs Agency paths, Key Security Indicators (KSIs), Marketplace listing, assessor or advisor obligations, agency sponsorship, continuous monitoring, significant change, vulnerability or incident timelines, FedRAMP public notices (NTC-), CISA BOD 26-04 for cloud providers, scope, OMB M-24-15, the FedRAMP Authorization Act — and also when someone asks how a SaaS or cloud product gets approved to sell to federal agencies, whether SOC 2 or GovRAMP counts toward FedRAMP, or what a "FedRAMP advisor" must do, even if the word FedRAMP is not used. Answers from the official FedRAMP 2026 markdown corpus and machine-readable rules bundled with this skill (auto-refreshed from GitHub and fedramp.gov), with FedRAMP notices and cited CISA/OMB documents as a labelled secondary tier, citing rule IDs and file paths, never from memory or third-party summaries.
- ---
- # FedRAMP Specialist
- This skill makes the official FedRAMP 2026 material the primary source for FedRAMP answers. It bundles two
- upstream repositories, a lookup tool over the machine-readable rules, and verified reference guides that
- say which page answers which question. A labelled secondary tier holds the few outside documents the corpus
- points to: FedRAMP's public notices, the CISA and OMB documents it links, and the legacy Rev5 playbook.
- **Why this matters.** FedRAMP was rebuilt in 2026: the Consolidated Rules replaced much of the older program,
- and they keep moving — six releases between 2026-06-25 and 2026-09-13 (`corpus/2026-markdown/changelog.md`). Most FedRAMP knowledge in circulation — including a model's own
- training data — describes the pre-2026 program. An answer that sounds right from memory is the most likely
- kind to be wrong. So every factual claim in a FedRAMP answer should come from the corpus and carry a
- citation someone can check.
- ## Where everything is
- All paths below are relative to this skill's directory (`~/.claude/skills/fedramp-specialist/`).
- | What | Path | Authoritative for |
- |---|---|---|
- | Narrative + generated rule pages (285) | `corpus/2026-markdown/` | explanations, context, audience guidance, how rules are presented |
- | Machine-readable rules | `corpus/rules/fedramp-consolidated-rules.json` | exact rule text, force words, affected parties, class variation, effective dates, change history |
- | Page map (generated) | `references/file-index.md` | finding the right page fast |
- | All rules, verbatim (generated) | `references/rules-digest.md` | seeing every rule at once |
- | Family table (generated) | `references/rule-families.md` | the 17 rule families, counts, dates |
- | Question routing + pitfalls | `references/routing-map-and-pitfalls.md` | **read this first** when unsure where to look |
- | Certification model | `references/certification-model.md` | 20x vs Rev5, Program vs Agency, Classes A–D, Class A, packages |
- | Dates and changes | `references/timeline-and-changes.md` | effective dates, deadlines, transition, release history |
- | Provider playbook | `references/provider-playbook.md` | the CSP journey and every ongoing obligation with its clock |
- | Agencies and FedRAMP | `references/agency-and-fedramp-roles.md` | agency use, sponsorship, agency and program duties |
- | Assessors, advisors, Marketplace | `references/assessors-advisors-marketplace.md` | recognition, IV&V, independence, advisor duties, listings |
- | Scope, law, definitions | `references/scope-authority-definitions.md` | what is in scope, statutory basis, M-24-15, key terms |
- | Where the corpus stops | `references/known-gaps.md` | contradictions, silences, errata — read before filling any gap |
- | Specialist briefing | `references/agent-briefing.md` | the mental model on one page (also embedded in the agent) |
- | Official JSON schemas | `corpus/schemas/` + `MANIFEST.json` | the 11 schemas the rules require documents to validate against |
- | Secondary sources (generated list) | `references/secondary-index.md` | what each outside document is, who cites it, and how far to trust it |
- | FedRAMP public notices | `corpus/secondary/notices/NTC-*.md` | official announcements the rules and narrative refer to (e.g. NTC-0014 on BOD 26-04) |
- | CISA / OMB documents the corpus links | `corpus/secondary/directives/` | the text of BOD 26-04 (its Table 1 is an image beside it), M-24-15, A-130, the MFA fact sheet |
- | Legacy Rev5 playbook (historical) | `corpus/legacy/content/playbook/` | the pre-2026 process the corpus still routes agency-sponsored Rev5 work to |
- The generated files are rebuilt from the corpus on every sync. The hand-researched guides were verified
- line by line against the 2026-09-13 rules and 2026-09-14 markdown; when a guide and the corpus disagree,
- **the corpus wins** — say so and use the corpus.
- ## Step 1 — Make sure the corpus is current
- ```bash
- bash ~/.claude/skills/fedramp-specialist/scripts/sync.sh
- ```
- This pulls the repositories and refetches the secondary documents only if the last sync is more than 7 days
- old, then prints the rules version and whether the markdown and JSON were built from the same commit. It is
- quick; run it at the start of any FedRAMP task.
- Run it with `--force` when the question is about **dates, deadlines, "current" or "latest" requirements,
- what changed recently, or anything a client will act on this week.** Those are the answers a stale snapshot
- gets wrong. If the network is unavailable, continue with the snapshot and state its version and date.
- Several agents can run it at once: syncs take turns under a lock, and `--force` skips a pull that finished
- in the last 10 minutes.
- If `sync.sh` warns that the markdown and JSON were built from different rules commits, prefer the JSON for
- exact rule text and say that the rendered pages may lag.
- ## Step 2 — Route the question
- 1. Classify it: which **party** (provider, agency, assessor, advisor, FedRAMP), which **track** (20x, Rev5,
- both), which **path** (Program, Agency), which **class** (A–D), and whether it needs exact rule text,
- narrative explanation, or both.
- 2. Open `references/routing-map-and-pitfalls.md` and find the closest question. It names the files and the
- command to run.
- 3. Check its pitfalls section if the question carries pre-2026 vocabulary (ATO, P-ATO, JAB, impact level,
- baseline, 3PAO, FedRAMP Ready). Answer in the corpus's current terms and say what the old term maps to,
- if the corpus says.
- 4. If nothing matches, search `references/file-index.md`, then grep the corpus.
- ## Step 3 — Get the facts from the source
- Use the lookup tool for anything a rule decides. It prints the record as stored, its effective dates, and
- the `path:line` where it renders, so the answer can cite both.
- ```bash
- R=~/.claude/skills/fedramp-specialist/scripts/rules.py
- python3 $R rule FRC-CLA-ASF MKT-CAS-WEB # full records
- python3 $R search "significant change" --affects Providers --force MUST --brief
- python3 $R search vulnerab --track 20x --class B # class- and track-filtered
- python3 $R family VDR # purpose, dates, pages, every rule
- python3 $R party Advisors --full # everything that binds a party
- python3 $R dates # effective dates for every family and track
- python3 $R define "significant change" # FRD definitions (binding inside rules)
- python3 $R ksi KSI-IAM # Key Security Indicators
- python3 $R control IR-4 # FedRAMP guidance on a Rev5 control (AC-2, ac-2.1, IA-2(1) all work)
- python3 $R changed --since 2026-08-01 # what moved, with FedRAMP's comment
- python3 $R info # version and provenance
- ```
- Then read the narrative page the tool points to (or the routing map names) for context. Rule pages come in
- three granularities: `reference/<family>.md` (everything), `reference/<20x|rev5>/<class>/<family>.md`
- (filtered to one track and class — use these when a class is named), and audience pages under
- `providers/`, `agencies/`, `assessors/`, `advisors/`.
- When a rule says a document must be valid against a FedRAMP schema, the schema is in `corpus/schemas/`
- (`MANIFEST.json` maps each to the rules that cite it). Validate the actual document against it rather than
- eyeballing field names.
- Three things are easy to get wrong, so check them explicitly:
- - **Force words are binding as written.** MUST, MUST NOT, SHOULD, SHOULD NOT, MAY each have an FRD
- definition. Never upgrade a SHOULD to a MUST or drop a MAY into "required".
- - **Defined terms change meaning.** When a capitalised term appears in a rule, its FRD definition is part of
- the rule even where everyday usage differs. Look it up before explaining the rule.
- - **Class applicability has two layers that sometimes disagree.** A subset's metadata can list Classes B–D
- while the rule itself carries a Class A variant, and Class A applicants must also address the specific rules
- listed in `FRC-CLA-MFR`. Read the rule-level statements, and check `references/known-gaps.md`.
- ## Step 4 — Answer
- - Lead with the answer, then the support.
- - Cite the most specific source for each claim: rule ID and/or `corpus/2026-markdown/<path>:<line>`.
- - Quote rule statements verbatim when the exact words decide the outcome.
- - Keep 20x vs Rev5, path, class, and party distinctions visible. If the answer differs by class, show the
- difference rather than averaging it.
- - Separate what the corpus says from inference, and label the inference.
- - State the rules version you used (`rules.py info`) when the answer involves dates or will be relied on.
- - If the corpus does not answer the question, say that plainly and say where it stops. Check
- `references/known-gaps.md` first — many silences and contradictions are already documented there, with a
- recommended way to answer meanwhile. Then check the secondary tier (below). Do not fill the gap
- from memory, and do not cite `fedramp.gov/docs`, blogs, vendor summaries or news as authority. Use those
- only if the user asks for history or public commentary, labelled as secondary.
- - A FedRAMP answer is guidance about the program's rules, not a legal opinion or a certification decision.
- Scope and eligibility calls are ultimately FedRAMP's (and the agency's) to make — say so where it matters.
- ## When the corpus points outside itself
- Some real questions turn on text the 2026 corpus names but does not contain. The commonest is **"what binds
- an existing Rev5 provider right now, before its 2026 dates?"**: the rules give obtain, maintain and grace
- dates, but they say little about the months before those dates. The secondary tier exists for these cases.
- Use it after the corpus, never instead of it, and keep the label visible in the answer.
- - **FedRAMP public notices** (`corpus/secondary/notices/`) are official. The corpus's own test is "If it is
- not written down in these rules or on fedramp.gov then it is not real" (`support.md:46`), and notices are
- on fedramp.gov. Cite them by number and date ("NTC-0014, 2026-06-16"). Notices can predate the final
- rules. If a notice and the rules JSON differ, give the JSON's value, mention the notice, and say which is
- later.
- - **CISA and OMB documents** (`corpus/secondary/directives/`) are authoritative for what they require of
- *agencies*. What a provider must do follows from the FedRAMP rules and notices that adopt them. For
- example, VDR-TFR-KEV points to BOD 26-04's timelines. Some figures in these documents carry content that
- is not in the text, such as BOD 26-04's Table 1. Open them with Read.
- - **The legacy playbook** (`corpus/legacy/`) is **historical**. FedRAMP's README for it says it is not a
- source of truth for current requirements. Use it only where the corpus sends people there
- (agency-sponsored Rev5 sponsorship and the legacy SAP/SAR, `agencies/sponsoring/initial.md:26`,
- `assessors/fedramp-assessments/rev5.md:42`), or to describe what an existing Rev5 authorization was built
- on. Always pair it with the dated 2026 rule that replaces it. Never present a legacy figure as a current
- FedRAMP rule. A client's own ATO and agency agreements may still impose it, and only the client can
- confirm that.
- `references/secondary-index.md` lists every document with its tier, source URL, fetch date and the corpus
- lines that cite it.
- ## When to hand off to the FedRAMP Specialist agent
- A subagent named `fedramp-specialist` carries the complete rule set in its instructions and works with the
- same tools. Delegate to it (Agent tool, `subagent_type: fedramp-specialist`) when the work is larger than an
- answer: a readiness or gap assessment against the rules, reviewing a client's package, SSP, or policies,
- drafting a client-facing deliverable, mapping a client's controls to KSIs, or checking a document for
- statements that contradict the 2026 rules. Give it the client facts and the files; it returns cited
- findings.
- ## Role boundaries
- The skill serves whoever is asking: a cloud provider, an agency, an assessor, or an advisor or consultant.
- Frame each answer for that party.
- - When supporting an advisor or consultant, keep advisory output clearly separate from anything that reads
- as an assessment result or a FedRAMP decision. Advisors prepare clients; they do not assess or certify.
- `references/assessors-advisors-marketplace.md` covers what the rules say about advisors and independence,
- and its advisor compliance checklist covers the `MKT-CAS-*` rules that bind a listed advisor, including
- the rule on answering FedRAMP and GSA email within a fixed number of business days.
- - The provider stays accountable for its package even when someone else helps prepare it.
- - Never promise a certification outcome or timeline that the rules and FedRAMP do not control.
- ## Keeping this skill current
- - `scripts/sync.sh` refreshes the repositories, the schemas and the secondary documents, then rebuilds
- `file-index.md`, `rule-families.md`, `secondary-index.md`, `rules-digest.md`, and reinstalls the
- `fedramp-specialist` agent so its rule knowledge moves too. When the secondary fetch reports a **NEW**
- notice, read it: notices are how FedRAMP announces changes between rule releases.
- - The researched guides are not regenerated. After a sync that reports changes, run
- `python3 scripts/rules.py changed --since <last sync date>` and check whether any changed rule is quoted in
- a guide; fix the guide if so.
- - `bash scripts/sync.sh --status` shows the snapshot age, versions and commits without changing anything.
README.md · README · 63 lines
- # FedRAMP Specialist (Claude Code skill + subagent)
- A Claude Code skill that answers FedRAMP questions from the official FedRAMP 2026 material and cites the
- rule or page behind each answer. It bundles FedRAMP's 2026 markdown corpus and machine-readable Consolidated Rules, a command-line
- lookup tool over the rules, and hand-researched guides that route each kind of question to the right page.
- A matching subagent carries every rule in its instructions for larger work such as gap assessments and
- package reviews.
- ## What's inside
- | Path | Contents |
- |---|---|
- | `SKILL.md` | The skill: when it triggers, how it routes questions, how it cites |
- | `agent-template.md` | Template for the `fedramp-specialist` subagent; the build fills in the full rule digest |
- | `corpus/2026-markdown/` | Clone of github.com/FedRAMP/2026-markdown |
- | `corpus/rules/` | Clone of github.com/FedRAMP/rules (`fedramp-consolidated-rules.json`) |
- | `corpus/schemas/` | The JSON schemas the rules name, fetched from fedramp.gov |
- | `corpus/secondary/` | FedRAMP public notices and the CISA/OMB documents the corpus cites (labelled secondary) |
- | `corpus/legacy/` | Pages-only clone of github.com/FedRAMP/docs-legacy (historical only) |
- | `references/` | Routing map, certification model, timelines, playbooks, known gaps, and generated indexes |
- | `scripts/sync.sh` | Refreshes everything above and rebuilds the indexes and the subagent |
- | `scripts/rules.py` | Lookup CLI: `rule`, `search`, `family`, `party`, `dates`, `define`, `ksi`, `control`, `changed`, `info` |
- | `evals/evals.json` | Three test prompts with pass criteria |
- ## Install
- Requires Claude Code, `git`, `bash`, and Python 3.9 or later (standard library only).
- ```bash
- git clone https://github.com/kamranarshad33/fedramp-specialist.git ~/.claude/skills/fedramp-specialist
- bash ~/.claude/skills/fedramp-specialist/scripts/sync.sh --force
- ```
- The repository does not include the three upstream FedRAMP repositories; the first sync clones them into
- `corpus/`. After that, the sync pulls the latest FedRAMP repositories, refetches the schemas and notices, rebuilds the indexes, and
- writes the subagent to `~/.claude/agents/fedramp-specialist.md` with paths for your machine. To write the
- subagent somewhere else, set `FEDRAMP_AGENT_PATH` first.
- ## Use
- - In Claude Code, ask any FedRAMP question, or call `/fedramp-specialist`.
- - For larger jobs, ask Claude to delegate to the `fedramp-specialist` subagent and give it the facts and file paths.
- - From a shell: `python3 ~/.claude/skills/fedramp-specialist/scripts/rules.py search "significant change" --brief`
- ## How it decides what is true
- 1. Rules JSON
- 2. 2026 markdown corpus
- 3. FedRAMP public notices (official, secondary)
- 4. CISA and OMB documents the corpus cites
- 5. Legacy playbook (historical only)
- The guides in `references/` were checked line by line against the September 2026 rules. When a guide and
- the corpus disagree, the corpus wins. FedRAMP changes its rules about once a month, so run `sync.sh` before
- any date-sensitive work.
- ## License
- The skill, scripts and guides are released under the MIT License. The FedRAMP, CISA and OMB documents in
- `corpus/` are U.S. government publications and are not covered by it. See `LICENSE` for the details.
- This project is not affiliated with or endorsed by FedRAMP or GSA.
- Its answers explain the program's rules; they are not legal advice or a certification decision.
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.