The fastest way to get a package bounced is not a weak control narrative. It is a validator opening the artifact list, seeing thirty files named Final_v3.pdf and scan2.pdf, and discovering that none of them are attached to a single control. I watched a Moderate package (call it SENTINEL, roughly 300 applicable controls) sit for eleven weeks because the team uploaded every document once at the system level and wrote “see SSP” in every implementation statement. The documents were fine. The package was unreadable. This post is the list of what a DoD authorization package actually needs, which control families each document proves, and the two artifact habits that cost new ISSOs the most calendar time.
An artifact is not evidence until it points at a control
An artifact is a file in the system record. Evidence is a specific place in that file (a section, a page, a screenshot, a log excerpt) that demonstrates a specific control, and really a specific assessment procedure under it, is implemented the way the narrative says it is. The SSP is an artifact. Section 4.3 of the SSP describing a fifteen-minute session lock, plus a screenshot of the group policy setting that enforces it, is evidence for AC-11.
The test I apply: a validator should be able to go from the control, to the attached artifact, to the page, to the fact, in under a minute. If they have to read the whole document to find the fact, you have given them an artifact, not evidence. A 140-page SSP attached to AC-2 is not evidence that accounts are managed. It is evidence that someone on the program owns Adobe Acrobat.
eMASS lets one artifact be attached to several controls, and that is the right behavior for a document like the SSP or the CM plan. The discipline is in the pointer. Every control that uses a shared artifact needs a section or page reference, either in the implementation narrative or in the artifact’s description text, so the reader lands on the fact instead of the cover page. The mechanics of attaching are in how to upload artifacts in eMASS; this post is about what to upload.
Building or renewing a package?
The ISSO’s ATO Package Kit is the 54-page guide plus the SSP template, 185 pass-ready control implementation statements, the artifact index and the pre-submission checklist validators use. Instant download, free updates.
How validators actually read a package
Nobody starts at AC-1 and reads forward. Every validator I have worked with opens the artifact list first, because it answers the first question: is this package real, or a shell with narratives in it? What they look at:
- Names and dates. SENTINEL_SSP_v2.1_2026-09-15.pdf tells them the system, the document, the version and the currency without opening it. Final_v3.pdf tells them you are going to be a long afternoon.
- Attachment pattern. Are artifacts attached to controls, or are they all floating at the system level? A system-level-only package means every control narrative is unsupported until proven otherwise.
- Internal consistency. Does the SSP date fall after the boundary diagram date and the hardware/software list date?
- The obvious gaps. No contingency plan test results. No signed appointment letters. No ISA for a connection that is plainly drawn on the diagram.
Then they sample the controls that are easiest to fake in a narrative and hardest to fake in evidence: AC-2 (account management), CM-6 (configuration settings), SI-2 (flaw remediation), CP-4 (contingency plan testing), IR-8 (incident response plan) and AU-6 (audit review). Last, they read the POA&M to see whether the failures in your own test results actually appear as items. If you want the full list of what sinks a package at this stage, the pre-submission rejection checklist is the companion to this post.
Free download
The ISSO’s eMASS Cheat Sheet (2 pages)
Access, workflow, control status, POA&M anatomy, inheritance, the narrative formula, and the ten mistakes that delay an ATO, on two printable pages. Written by a working ISSO, not a vendor. Enter your email and it’s yours. The full package build, templates included, is in the ATO Package Kit.
Get the free cheat sheetThe core documents, what each one proves, and how each one goes wrong
Family references are NIST SP 800-53 Rev 5. Your component and your AO may add to this list; they rarely subtract from it.
Foundation documents
- System Security Plan (SSP). Proves PL-2 directly and is the parent of every control narrative, so it is referenced by every family. The mistake: attaching the whole SSP to dozens of controls with no section reference, and letting it drift from the diagram and the hardware list it is supposed to describe.
- Authorization boundary diagram. PL-2, PL-8, SC-7, CA-3, AC-4. The mistake: drawing a network diagram instead of a boundary. A boundary diagram shows what is inside, what is inherited, every external connection and the flows across the line. I wrote up what validators want in the authorization boundary diagram post.
- Hardware and software list. CM-8, CM-2, CM-7, SA-22 for unsupported components, and it sets the scope for RA-5. The mistake: exporting it once at registration and never again, so the asset count disagrees with the ACAS scan count. The hardware/software baseline post covers how to keep those two numbers matched.
- Signed appointment letters (ISSM, ISSO, and where applicable system owner and AO designation). PM-2, PL-2 roles, CA-6. The mistake: unsigned letters, or letters signed for a person who left the program two contracts ago.
Risk and assessment documents
- Risk assessment report. RA-3, RA-2, and the context for RA-5 and CA-5. The mistake: relabeling the categorization memo as a risk assessment. A risk assessment names threats, likelihood, impact and the residual risk the AO is being asked to accept.
- Security assessment plan and security assessment report. CA-2, CA-7, and the SAR feeds CA-5. The mistake: findings in the SAR that never become POA&M items. Validators cross-check these two documents line by line.
- POA&M. CA-5 directly, and it is the closure record for failures under RA-5, SI-2 and CM-6. The mistake: entering it once and never touching the milestones. A POA&M whose scheduled completion date was last March is not a plan. It is a confession. Mechanics are in how to enter a POA&M in eMASS.
- Privacy artifacts (privacy threshold analysis, PIA, SORN reference) when applicable. The PT family, plus SI-12 and the privacy portions of PL-2. The mistake: marking PII as not applicable with nothing to back it. Even “we hold no PII” needs a signed threshold analysis that says so.
Operational plans
- Contingency plan and test results. CP-2, CP-3, CP-4, and the recovery controls CP-6, CP-7, CP-9 and CP-10. The mistake: a plan with no test, or test results that are a sign-in sheet. CP-4 wants what was exercised, what failed, and what changed in the plan because of it.
- Incident response plan. IR-1 through IR-8, with IR-4, IR-6 and IR-8 carrying the weight, plus the AU-6 reporting link. The mistake: copying the enclave or site IRP with no system-specific contact roster, escalation path or reporting timeline.
- Configuration management plan. CM-1, CM-3, CM-4, CM-5, CM-9 and SA-10. The mistake: a plan that describes a change control board with no board minutes, change tickets or approved baselines attached anywhere in the package.
- Continuous monitoring strategy. CA-7, PM-31, SI-4, RA-5, AU-6. The mistake: a strategy that says monthly scanning while the SLCM fields on the controls say annual. Validators read both.
- Interconnection security agreements and MOUs/MOAs. CA-3, AC-20, SC-7, SA-9. The mistake: expired agreements, or a connection drawn on the boundary diagram with no agreement behind it.
Evidence files
- STIG checklists and ACAS results. CM-6, CM-7, SI-2, RA-5, IA-5, AC-17 and whichever technical controls the applicable STIGs map to. The mistake: uploading one CKL per host as loose artifacts, stale relative to your own monitoring frequency, with no summary that ties open findings to POA&M items.
- Account management evidence. AC-2, AC-3, AC-5, AC-6, IA-2, IA-4, PS-4, PS-5. The mistake: exporting the user list and calling it done. The evidence is the access request, the approval, the signed periodic review, and the disabled account that proves termination ran.
- Training records. AT-2, AT-3, AT-4, PS-6. The mistake: a screenshot of one person’s cyber awareness certificate. Validators want a roster that reconciles against the account list, including privileged user training for everyone with elevated access.
The quick mapping, artifact to family
- SSP: PL (and referenced by all)
- Boundary diagram: PL, SC, CA, AC
- Hardware/software list: CM, SA, RA
- Appointment letters: PM, PL, CA
- Risk assessment report: RA, CA
- SAP and SAR: CA
- POA&M: CA, closes RA, SI, CM findings
- Privacy artifacts: PT, SI, PL
- Contingency plan and test: CP
- Incident response plan: IR, AU
- Configuration management plan: CM, SA
- Continuous monitoring strategy: CA, PM, SI, RA, AU
- ISAs and MOUs: CA, AC, SC, SA
- STIG and ACAS results: CM, SI, RA, IA, AC
- Account management evidence: AC, IA, PS
- Training records: AT, PS
Sixteen documents is the floor, not the ceiling. The artifact index in the ATO Package Kit does this mapping for 69 artifacts and lists the questions a validator will ask about each one, and it is what I use when scoping a new package.
The two attachment mistakes that cost the most time
New ISSOs fail in one of two opposite directions. I have done both.
Mistake one: upload everything once at the system level and never attach it to a control. The package looks complete from the artifact list and empty from every control view. The validator opens AC-2, sees a narrative and nothing attached, and marks it unsupported. Multiply by 300. The fix is tedious: every control marked compliant gets at least one attached artifact with a pointer to where the evidence lives. System-level is for things that are genuinely system-wide and not control-specific: the authorization letter, appointment letters, the ISA binder.
Mistake two: attach the same PDF to forty controls with no pointer. This is what the same team does after being bounced for mistake one. The SSP now appears under forty controls and the validator opens forty copies of the cover page. The fix is the pointer: a section number or page in the narrative or in the artifact description. For the controls that get sampled hardest, attach a short extract or screenshot beside the parent document so the fact is on the first page.
When your program has no document for a required item
This happens on every first-time package I have touched: no contingency plan, a CM process that lives in one engineer’s head, no ISA for a connection that has existed for six years. Four options, in the order I try them.
- Inherit it. If the hosting enclave or the common control provider owns the function, the artifact is theirs and you inherit the control. Confirm the provider’s package actually covers your system first.
- Write the minimum honest version. A four-page contingency plan that is true beats a forty-page one copied from another system.
- Capture the process that already exists. If changes are approved in a weekly meeting with email records, the CM plan is a description of that meeting plus the emails as evidence.
- POA&M it. Mark the control other than satisfied, open a POA&M item with a real milestone and owner, and say so in the narrative. An AO can accept a documented gap. An AO cannot accept a gap you hid.
What you never do is mark the control compliant and attach a document from another system with the name changed. Validators have usually read that document before, under its original name.
The package that moves is the one where the artifact list reads like an index, every compliant control points at a fact, and the gaps are declared instead of buried. Build the artifact list first, map each document to its families before you upload anything, and the control narratives almost write themselves, especially if you start from the ATO Package Kit’s artifact index and SSP template instead of from a blank page.


Leave a Reply