Repository navigation
Description
A few updates:
Party Table
Mapping providers :
REPLACE CURRENT WITH:
" The mapping provider publishes framework mappings for catalogs published by the regulatory control providers. The mapping provider is NOT involved in mapping controls to product recommendations for implementing or satisfying the controls. The mapping providers' OSCAL activity is focused on ONE OSCAL artifact: the mapping between regulatory frameworks"
Technology providers
REPLACE CURRENT WITH:
: The technology provider publishes the component definition for their product hardening recommendations/rules (in English) on implementing or satisfying the controls. The technology provider publishes in the component definition for each rule the evidence type (url/id) their APIs or documentation corpus makes available to collect as proof. The technology provider of policy validation engines publishes the component definition for their engine's checks (ids) that implement the hardening rules of their targeted assessed products. Only the rules products components carry the tie to the framework, not the associated policy validation engines components. The technology providers' OSCAL activity is focused on ONE OSCAL artifact: the component definition with their product knowledge.
Policy engine providers
REPLACE CURRENT WITH:
The technology provider of policy validation engines publishes the component definition for their engine's checks (ids) that implement the hardening rules of their targeted assessed products. The technology providers' OSCAL activity is focused on ONE OSCAL artifact: the component definition with their product knowledge.
System owners
REPLACE CURRENT WITH:
The system owner (1) selects regulations or merges catalogs through an SSP imported profile; (2) selects the components (defining hardening rules and evidence types) associated with their products inventory in the SSP system implementation; (3) selects the assessment assets components in the Assessment Plan thus deciding on policy engines pre-defined checks to be performed ; (4) adds new checks to the Assessment Plan specific to their systems (SEE Assessment First approach). Using OSCAL mapping artifacts, the SSP and Assessment Plan can be transposed to mapped regulations.
Independent auditors
REPLACE CURRENT WITH:
The auditor's run produces assessment results: (1) Observations results at assessment asset check level per inventory item, which are aggregated via the components checks -> rules -> controls association to (2) Findings results at target control level.
REMOVE or ADD to ALL
Should assessment platform usage be optional in an assessment plan?
when is an assessment platform also needed?
How should an assessment asset be associated with tasks or activities?
How should an observation identify the assessment asset that produced it when several assets are declared?
Comment: These are OSCAL generic question not associated with CDef First, we can add it to all three approaches
REMOVE When is an assessment component sufficient,
This can be asked of any element : when is a catalog list of controls sufficient?
GOOD QUESTION to ADD to CATALOG FIRST
How should an SSP combine benchmark and vendor CATALOG that describe the same system recommendations ?
NEED UPDATE
How should an observation identify the assessment asset that produced it when several assets are declared?
NOT ACCURATE, REMOVE AND REPLACE WITH:
An assessment result may declare observations from several assessment assets. The assessment asset id is provided in the Observation augmented schema, establishing unambiguously the relationship between an observation identify and the assessment asset that produced it. The component-first complete path from Assessment Results observation to producing assessment asset to associated control is as follows: Observation id; assessment asset inventory id; assessment subject inventory id; component types associated with them in the Assessment Plan and SSP; check -> rule -> control associated with them in the Component Definition; transpose to other regulation via Mappings.
ALL the remaining questions need to be reviewed and answered from the perspective of the CDef First model. BRB
Activity
ancatri commentedon Sep 21, 2026
Added all QA for reference
The rule sits on the component it configures. Open questions concern runner identification, importing component definitions and tailoring benchmark content.
Which component definition carries the tie to the framework, the one holding the rules or the one holding the checks?
The control link is inline rather than held in a mapping collection. The software component definition holds rules, while the validation component definition holds checks; those artifacts may have different maintainers. A mapping provider could update either or both. The OSCAL examples do not establish which document should own that relationship or how duplicated mappings would stay consistent.
Should assessment platform usage be optional in an assessment plan?
An assessment plan may describe a runner as a component, a platform using components, or a tool identified through properties. These alternatives leave open when an explicit assessment platform adds necessary information and when a component reference is sufficient for a consumer to execute or interpret a check.
When is an assessment component sufficient, and when is an assessment platform also needed?
The OSCAL examples use different combinations of assessment-assets, assessment-platforms and components. They do not establish a shared rule for deciding whether a component alone identifies the runner or a platform must also describe its execution context. A consistent distinction would help consumers interpret equivalent workflows.
How should an assessment asset be associated with tasks or activities?
A task identifies work to perform, while an assessment asset describes resources available to perform it. The component-first examples do not establish a complete, interoperable association between these structures. The question is which references should identify the runner for a particular task or activity.
How should an observation identify the assessment asset that produced it when several assets are declared?
An assessment result may declare several assets, so naming an observation's producer requires an unambiguous relationship. The component-first result example does not establish a complete path from observation to producing asset. The open question is how consumers should recover that association without relying on tool-specific assumptions.
How should an SSP combine benchmark and vendor component definitions that describe the same system component?
A system may use vendor component definitions alongside benchmark component definitions for the same software. Importing both can produce overlapping descriptions, rules and control implementations. It remains open whether an SSP should retain multiple OSCAL components, merge them, or use another convention to represent the single system component consistently.
Should desired-state values from benchmark component definitions be replaced with actual-state values in the SSP?
Rules describe prescribed configuration, while an SSP records the system owner's implementation claims. Replacing prescribed values with actual values could lose the benchmark baseline; retaining only prescribed values could misrepresent implementation. The question is how to keep the two meanings distinct while preserving the relationship between benchmark requirements and the system.
Which benchmark context should enter an SSP, and how should tools filter component-definition content during import?
Benchmark content can include rules, parameters, rationale and procedural detail, not all of which serves the same purpose in an SSP. Import tooling needs a policy for retaining or omitting that context. The open issue is how to reduce duplication without losing information needed to interpret an implementation claim.
How should component definitions support benchmark selection, exclusions and parameter tailoring?
The component-first examples place parameters on proposed rules and set-parameters on control implementations. Those structures address parameter values but do not provide a profile's selection and exclusion workflow. How component definitions should support benchmark tailoring, including the scope and persistence of tailoring decisions, remains open.
Should assessment plans or assessment results support importing component definitions with benchmark context?
The component-first examples carry check content into an assessment plan through assessment assets. That demonstrates one encoding, not a dedicated import mechanism for all benchmark context. The question is whether plans or results should support such imports directly and which rules, parameters and provenance would need to survive.
ancatri commentedon Sep 22, 2026
On the open questions section https://oscal-foundation.github.io/Pattern-Library/docs/analysis/2026-08-automating-technical-hardening-guidance/questions.html
The distribution and framing of open questions across the three approaches warrants scrutiny.
The asymmetry in depth, placement of generic questions, and choice of framing language produces a presentation that does not treat the three approaches on equal analytical footing, regardless of whether that was the intent.
To-dos to rebalance and clarify
Rebalance question distribution
Audit every CDef question for whether it applies equally to Catalog-first or Assessment-first; move generic questions to Section 1 (Across all three approaches)
Add equivalent questions to Catalog-first on SSP desired-state vs actual-state and overlapping component definitions
Expand Assessment-first questions — one question for the least conventional approach is not credible; at minimum address runner identification, evidence traceability, and reuse of check content across assessments
Reframe CDef questions
Rewrite questions framed as structural ambiguities ("which document should own," "how should consumers recover") to match the maturity/scale framing used for Catalog-first
Replace "remains open" language where a reasonable convention exists in published OSCAL examples — don't leave CDef questions open when equivalent catalog-first questions get a worked answer
Clarify CDef approach positively
Add a worked CDef scenario equivalent in depth to the catalog-first scenario — if one approach has a worked example and others don't, the document is not comparative
State explicitly what CDef solves that the other two don't: product-owned rules, component lifecycle independence, tooling-native representation
Document the runner identification pattern as a convention, not an open question — OSCAL Compass has implemented this; it is not unresolved
Structural fix
Add a preamble to the open questions section stating the editorial standard: questions are placed under the approach where they are most acute, not exclusively — generic questions belong in Section 1
Request authorship transparency: who wrote the questions section and was it reviewed by a proponent of each approach before publishing
vikas-agarwal76 commentedon Sep 24, 2026
@pjavan I posted this message to you and Anca on Slack channel on Aug 19 for updating the answered/un-answered for the six questions on the main page. Capturing it here so that these can be incorporated in the updates .....
I went through the index.html file that compares the 3 approaches. This looks very good to me.
For our proposal of using cDef, I believe few aspects have not been captured accurately as I may have not included every detail in the examples or the markdown file that I shared earlier. But, based on your 6 questions for comparison, I am including here the changes that need to be made -
2b. Can a third party link this rule to a control after publication, without touching the rule? - answered
The cDef is provided by the vendor, but the SSP is the one that gives the actual implementation. The association of rules to controls which is there in cDef -> implemented-requirements can always be changed/overridden in the SSP by the implementor. Also, if there is no association of rules to controls in cDef, the implementer can provide their own association in SSP.
4. What the check is run against - answered
In cDef the type of the component tells which component is the rule run against. Everything - system, software, process, etc is a component in SSP. Even the whole env. is the default “This System” in SSP. So the approach answers this question on what the check is run against.
6a. How does a system say it satisfies this, in its own words? - answered
Each rule has a description that gives the text of what the rule does and how is it satisfied.
Also, when a rule is associated with the control in implemented requirements, the existing description field can be used to specify how that control is implemented, in addition to associating the rules with the controls. I have added text in the examples to explicitly capture this.
Also, updated this in SSP. And added SSP example in the markdown file.
6b. Was it satisfied, when, and on what? - answered
Didn’t include the findings in the assessment result as there were no changes proposed to it. I have added the findings in the AR example to capture how the observation for a rule is associated to a control statement in findings. Also, if no findings is present in the AR, one can always compute it starting from the observation using the association check-id -> rule-id -> control-id. These associations are present in the 2 cDefs.
pjavan commentedon Sep 24, 2026
@ancatri and @vikas-agarwal76. I believe accounted for most, but likely missed something. I've opened PR#17 to track these changes.
Adjustments I am making
I am making this adjustment.
I am making this adjustment.
I am making this adjustment.
I am making this adjustment.
Summary: The desired-state vs actual-state question applies to any approach that produces SSP content, not only component-first.
I am making this adjustment
Comments I am not adjusting for
I am not making this adjustment because I'm not tracking with Observation augmented schema. I understand an observation, risk, or finding represents the assessment asset used through the origin->actor->type. We have this capability in AR, but I do not see a way to do this in an AP.
I am not fully making this adjustment because I'm not sure I'm tracking. Assessment-first has related-evidence (evidence traceability) in AR, not possible to be captured in the SSP. I've added this destinction. runner identification is the assessment platform right? I believe we capture that and how it can (in AR) or not (in AP) map back to specific tasks. I am not trackign what "reuse of check content across assessments" means
I am not making this adjustment because I'm not sure what to do. If there are specific component-first questions that you would like written differently please post back and we'll make the updates.
I am not making this adjustment because I took what was available. If there are more examples, but to clarify I'm assuming you mean the examples here: https://oscal-foundation.github.io/Pattern-Library/docs/analysis/2026-08-automating-technical-hardening-guidance/oscal-artifacts.html. These were sourced and not generated. The six questions normalized on two requirements: data at rest and Minimum password length
I am not making this adjustment because I'm not sure how the cdef approach is a better tooling-native representation. Does the model influence how well the tool is represented? component life cycle independence wasn't a goal of this effort and produce-owned rules I believe are supported in both catalog and component first. AP-first could benefit from this, but would need the Cdef/Catalog first to be where the rule (assuming that's what you mean by product owned) is defined
vikas-agarwal76 commentedon Sep 25, 2026
Hi @pjavan Thanks for your detailed response. For some of the changes that you are not planning to make because of lack of clarity, here are some further explanations. We can also discuss this in the call as that will help further clarify these -
Comments I am not adjusting for
The main question was - "How should an observation identify the assessment asset that produced it when several assets are declared?"
This was one of the questions we listed for the current OSCAL AR schema and then provided a solution for it by augmenting the AR schema with assessment asset uuid and check id in each observation. Also, this question is applicable for all approaches and not just cDef first.
May be I am not understanding the objection here properly, but I don't understand "not tracking the Observation augmented schema...." means. Since the question is about "observation" the response was about changes in the AR schema which describes how each observation can be associated with the assessment asset uuid.
If the question's intent was about AP then please reframe the question accordingly. But, as discussed in our automated assessment meetings earlier, even for AP we discussed how the assessment asset is associated with check ids. The AP example also captures this.
Here as you see we added the "assessment-asset-uuid" and the "check-ids" to activities in AP. Hope this clarifies this.
The are few open questions in Assessment first approach (and prob. Catalog first also) like -
vikas-agarwal76 commentedon Sep 25, 2026
@pjavan Here are answers to some of the questions listed for cDef approach.
1. How should an SSP combine benchmark and vendor component definitions that describe the same system component?
Currently there is no support in OSCAL for importing a cDef into other models such as SSP, AP, etc. They need to be copied and manually customized while putting it into SSP, AP, etc.
In the cDef first approach we are following this same mechanism for including cDefs in other models. This also enables one to tailor/customize the cDef while copying. For example, if the cDef for a component defines 8 chars as default value for a password, while copying into SSP the platform owner can use any other value based on the requirements. Similarly, a cDef may define many rules, but while including into SSP one can only include those rules that are relevant based on the profile controls for that SSP. Hence, the tailoring of cDef becomes part of the process of including it in other models.
This same approach can be used if there are 2 or more cDefs for same component by different providers. The SSP author can tailor those and create one cDef while including it in the SSP. Ultimately, what goes into SSP is decided by the SSP author and they can modify existing cDefs while copying into SSP.
The idea here that the cDefs provided by the vendors are library (or class) of components and while including in the SSP it becomes an actual instance (customized) for that SSP.
If we need support for importing cDefs (not just copying) then we would need another model similar to profiles for tailoring of cDefs. We haven't suggested that currently, but in future if the community feels the need then that can be done.
2. Should desired-state values from benchmark component definitions be replaced with actual-state values in the SSP?
As described above, the SSP author can set the values of parameters in SSP for cDef rules while copying the cDef into SSP. Also, since the cDef is not imported but copied, the SSP author will need to set some value (as there is no linkage to the original cDef). It is perfectly fine to either use same, or more stringent, or less stringent value based on the control requirements or organizational policy requirements. The ones provided in the original cDef are what the product vendor had as default, which can be tailored by the SSP author while copying into SSP.
3. Which benchmark context should enter an SSP, and how should tools filter component-definition content during import?
This situation is same as current. Even today when an SSP author copies the cDef into SSP, they need to decide what needs to be copied (all or some). Same logic is to be followed. Some information is required like - rule-id, description, parameter-id and its value, association of rules to controls using implementing-rules. Rest like audit and remediation procedure etc. are optional. It is upto the SSP author to decide whether to include or not.
If there was a cDef import mechanism then this duplication would not be required, but this is same as today. In future, if OSCAL supports cDef import and tailoring mechanism then this duplication can be avoided.
4. How should component definitions support benchmark selection, exclusions and parameter tailoring?
Already discussed the tailoring process in above questions.
5. Should assessment plans or assessment results support importing component definitions with benchmark context?
I am assuming benchmark context here refers to rules (and not checks). Checks details (and their association to rules) are directly included as part of the validation component in AP through copying of the content into assessment-assets in AP.
The benchmarks (i.e., rules) are included in the SSP only (and associated to controls through implementing-rules). They flow to AP via import of SSP into AP and into AR via import of AP into AR. Hence, benchmarks content are not required to be imported (or copied) separately into AP or AR. Any tailoring done in SSP will be available to AP and AR.
vikas-agarwal76 commentedon Oct 5, 2026
@pjavan @ancatri I have created 2 new cDef files -
one for "COS Service" + its ansible validation component in one cDef file, and
another for "idservice" + its oscap validation component in one cDef file.
This will make it easier to look for a software/service cDef and its corresponding validation cDef in one place. The contents are same as the 4 individual cDef files. Kindly add these in the OSCAL patterns repo.
idservice-oscap-validation-cd.json
cos-ansible-validation-cd.json
pjavan commentedon Oct 9, 2026
Thanks @vikas-agarwal76, having each service and its validation component in one file makes the pairs much easier to follow. I've added both files to
examples/component-first/in #17, with your file names:idservice-oscap-validation-cd.jsoncos-ansible-validation-cd.jsonI compared both files with the four separate ones, and every component matches exactly. I kept the four separate files too, because parts of the analysis quote them directly. The combined files are an easier way into the same examples, not a replacement.
One change I made: new document UUIDs. Both combined files came in with the document UUID
bbdf7ee6-3e0a-4065-8ecf-50f4129de473, which all four separate files already use. To avoid that collision, I gave each combined file its own UUID and updatedlast-modified. I didn't change anything inside the components.idservice-oscap-validation-cd.json5387e03f-f015-4c6c-a2b8-384f2319c087cos-ansible-validation-cd.json51f93843-4b96-4794-a590-1cbc6bad9b73One concern for you to check: the OSCAP checks point at a component that doesn't exist. In
idservice-oscap-validation-cd.json, all three OSCAP checks have:The IDSERVICE component's UUID is
0c410eb1-52cd-40bc-bf1d-c80126a984e9, so no component in the file matches...984f9. The separateoscap-validation-component-definition.jsonalready has this value, so it isn't new. It's just easier to see now that both components are in one file. The assessment plan uses...984e9, so it looks like a typo. The COS/Ansible file is fine because its checks point at the COS component.I haven't changed this value because the checks and their targets are your content. Can you confirm whether
...984e9is the intended target? If it is, I'll correct it in both the combined file and the separate one.pjavan commentedon Oct 9, 2026
@ancatri @vikas-agarwal76 I went back through this whole thread and compared it with the site. The remaining changes are now in #17:
Shared questions
Component-first
The party table now follows Anca's role descriptions:
The notes in the table had to be short, so the full wording didn't fit.
Since the rule component carries the tie, the "which component definition carries the tie" question now asks what's still open: how a framework the vendor didn't map can reach the same rules, either through an SSP override or a mapping between frameworks.
Questions on combining component definitions, benchmark context in the SSP, tailoring, and import into plans and results now describe Vikas's copy-and-tailor approach. They stay open on what it leaves unresolved: tracking which source each merged rule came from, duplicated content, keeping tailoring auditable, and a dedicated import mechanism.
Comparison table: question 4 is now answered for component-first, using Vikas's explanation of SSP components and "This System". Question 5 now describes how a runner is identified: each plan activity names its asset through
assessment-asset-uuidand its checks throughcheck-ids.Still open, and what we need
Alternatively, you can commit changes to the PR as well
vikas-agarwal76 commentedon Oct 11, 2026
Thanks @pjavan for catching this typo. Yes,
...984e9is the right id to use.vikas-agarwal76 commentedon Oct 11, 2026
@pjavan
Please find the SSP and the AR files here -
ssp.json
assessment-result.json
The current proposal allows association of product's technical rules to controls through implementing-rules. Since tailoring of cDefs are currently not supported in OSCAL, it can only be done by copying and editing. We can propose a tailoring mechanism for cDefs (similar to profiles) but that will be a separate discussion altogether. The cDefs can be tailored in the SSP by copying and editing as described earlier.
If a third party (and not system owner) wants to tailor the cDef provided by product vendor, they can follow the same approach by creating a new cDef for the same component and copying and editing (changing association of rules to controls) and publishing it. So now, there will be 2 cDefs available for the same product, the original one by product vendor and another by a third party. SSP authors can decide to use any one of them and also tailor them while including in their SSP.