attest-it
Skill2 starsChanged 21 days agoMITView on GitHub
Prove a codebase or change-set conforms to a standard — EU AI Act Art-12/50, SOC 2, NIST SSDF — with auditor-grade, independently re-derived evidence, or name the gaps. The unit is one obligation from a FROZEN standard catalog (standard@version), not a discovered finding: enumerate the obligations, gather evidence bound to authoritative state for each, have an independent session re-derive it, and mark each control VERIFIED or GAP. Use when "prove compliance", "conformance", "audit-ready evidence", "SOC 2 / EU AI Act / NIST SSDF attestation", "regulatory audit trail". Not for a security exploit loop (harden-it), a PR merge verdict (review-it), or closing a discovered backlog (clean-sweep).
In this repository22 other files it gives its agents.
- AGENTS.mdAGENTS.md
- absorb-itSkill
- access-itSkill
- clean-sweepSkill
- deflake-itSkill
- document-itSkill
- field-test-itSkill
- floor-itSkill
- harden-itSkill
- map-itSkill
- migrate-itSkill
- modernize-itSkill
- offload-itSkill
- oncall-itSkill
- oss-contributeSkill
- pin-itSkill
- prove-itSkill
- reshape-itSkill
- review-itSkill
- root-causeSkill
- ship-itSkill
- speed-itSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
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
- Very specific: line 77
Every obligation in the frozen standard@version catalog is accounted for: the union of unit contracts equals the standard's obligation set …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
attest-it — evidence-bound conformance to a frozen standard
You are the COORDINATOR of an evidence-bound conformance run. "Prove this conforms to standard X,
or name the gaps" is a user-facing outcome AND an audit artifact: each control is either satisfied with
evidence an independent session re-derives, or it is a named GAP parked to a human/legal owner — never
a silent pass. Composes decide-and-freeze (FREEZE is the one-way gate), decompose-dag, acceptance-review, research-brief (catalog sourcing),
completion-audit (an EXTERNAL-STATE control is a GAP, never an assumed pass), human-handoff
(every GAP is filed with an artifact, an owner, and a verify-complete observation),
publish-evidence (the attestation bundle leaves the tree as a versioned link); rides
evidence-manifest (each control's
evidence binds to authoritative state via the Art-12/50 provenance block and is re-derived, not
narrated), gate-classification (a GAP that needs a policy/spend/legal decision is a one-way human
door), sandbox-policy (PROFILE=ro for evidence gathering), threat-model (the fleet's own
trust assumptions and four boundaries — the scoping input when the system under assessment is a
software factory), ledger-contract (the obligation
ledger), liveness-resume (DAG-scoped RESUME). Worker TASK pack: one of matt | addy | gstack — never co-mount.
Terminal outcomes
- CONFORMANT — every obligation in the frozen catalog is VERIFIED: evidence bound to authoritative state, independently re-derived, no control marked satisfied without a linked artifact.
- CONFORMANT-WITH-GAPS (degraded) — one or more obligations are GAPs, each parked to a named human/legal owner with the missing evidence stated. A GAP is a first-class terminal, not a failure to hide.
Pipeline
REFUSE-IF-QUARANTINED (runtime/scripts/preflight.py --mode readonly --mission attest-it) → FREEZE (decide-and-freeze, ONE-WAY) the
denominator: standard@version and its SOURCE — the standard's own text or a human-supplied obligation catalog, as a
file/URL at a digest; obligations are enumerated FROM that document, never from model memory (a recalled SOC 2 / SSDF
control list is a hallucinated denominator) — then enumerate into a DAG (decompose-dag). The human confirms the
catalog; a catalog the human already named at a digest is VALIDATEd, not re-grilled; headless/spawned with neither
PARKS. The set is frozen at the digest — the run cannot shrink it.
→ EVIDENCE (per obligation, ro workers): gather the artifact that satisfies the control (a log field,
a policy version, a test outcome, a scan result), bound to authoritative state in an evidence
manifest with the provenance block (spec_version · model · reviewer · retention · standard).
→ RE-DERIVE (independent session, acceptance-review): a different session confirms each control's
evidence against authoritative state — deterministically where possible (verify.py), never the
gatherer's own narration. The re-derivation confirms each control's `provenance.standard` +
`spec_version` EQUAL the FROZEN catalog's `standard@version` (verify.py's `check_provenance`
enforces the Art-12/50 audit fields are present); a control evidenced against a different version,
or with no re-derivable artifact, is a GAP, not a pass.
→ AGGREGATE: every obligation VERIFIED or GAP; GAPs carry the missing evidence + a human/legal owner.
→ PUBLISH the attestation bundle as a versioned link (`publish-evidence`).
→ VERDICT: CONFORMANT, or CONFORMANT-WITH-GAPS with the gap register.Convergence proof (definition of done)
Every obligation in the frozen standard@version catalog is accounted for: the union of unit contracts
equals the standard's obligation set (no obligation unassigned), each VERIFIED control carries evidence
bound to authoritative state and independently re-derived (not the gatherer's narration), and each GAP
is named with its missing evidence and a human/legal owner. The verdict is CONFORMANT or
CONFORMANT-WITH-GAPS; advancing past a degraded terminal (accepting residual gaps) is a one-way human
gate. No obligation is silently dropped.
Ledger + supervision
Ledger header at T0 — the full canonical liveness-resume.md header
(RUN · COORDINATOR · BASE · FORK_POINT · T0 · SOURCE: <frozen catalog ref+digest> · WIP: builders=<n> reviewers=<n>) —
plus one row per obligation (obligation-id → task → VERIFIED | GAP), written before dispatch so an
interrupted run is recoverable, not a resume orphan (RUN / COORDINATOR are what RESUME dies
without). Stalls / death / compaction → liveness-resume.md (death → RESUME; compaction → write the
CONTEXT HANDOFF block, then RESUME): RESUME's scope is that header + the obligation rows,
git-verified — it reconstructs the DAG and re-enumerates which obligations were dispatched, VERIFIED,
or GAP, so the convergence verdict is never trusted from memory.
Anti-patterns
Marking a control satisfied on the gatherer's word (evidence must be re-derived independently — the
whole point). Shrinking the obligation set to the ones you can pass (the denominator is frozen at a
version digest). Treating a GAP as a failure to bury instead of a named, parked terminal. Confusing
attest-it with harden-it (attest proves conformance to a standard; harden runs an exploit→fix→re-attack
loop) or review-it (a per-PR merge verdict, not a standard's obligation set). Re-using an evidence
artifact whose provenance.spec_version no longer matches the frozen standard version. Enumerating
obligations from model memory (the catalog is a sourced document at a digest; a plausible control list
is not the standard).
Related
review-it (a per-PR verdict, not a standard's obligations), harden-it (the security exploit loop),
clean-sweep (close a discovered backlog); evidence-manifest (the Art-12/50 provenance block attest-it
attests against), mission-scheduling (report-heavy, schedules cleanly as an unattended conformance sweep).
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
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.

