F02: Assessment report and package (ADR 0014, 0015); register agents in KAB
Some checks failed
check agents / check (push) Failing after 14s

- Report (cp1-assessment-report-agent 0.2.0, step s7): a model writes the
  Assessment report from facts worked out in code (register, gaps, scorecard,
  rule traces, checks, and what each application is from the Discovery
  package); code sets the verdict and renders the pages. The
  disposition_signoff review point moves from the critic to the report.
- Package (cp1-assessment-package-agent 0.1.1, step s8): after sign-off, every
  step's output in one package, with a register, a gaps register, the
  sign-off record, lineage and a sha256 per part, in code. Page step s9.
- scripts/kab_register_agents.py: registers deployed agents in KAB's agent
  catalog (an in-place workflow update doesn't).
- Runs exec_1b1257380df7 (014) and exec_374d5d577077 (015) completed on
  crucible-kab-dev: critic pass, report figures all in its facts, package
  complete with every digest recomputing.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
2026-11-08 20:19:10 +00:00
parent 3b3ca64de0
commit 66c32745aa
58 changed files with 7106 additions and 26 deletions

View File

@@ -3,14 +3,18 @@
The same export format as the ALZ-F01 individual workflow (export_version 2.0.0). There is no orchestrator: each
step is one agent, and each step's inputs are run inputs or named fields of earlier steps' outputs (`from_steps`).
Profiling, risk and security posture run in parallel; the critic's step stops for the Engagement Lead's approval,
which is the `disposition_signoff` gate (ADR 0004). The build refuses a critic on one of its executors' models (K-12, ADR
Profiling, risk and security posture run in parallel; the report's step stops for the Engagement Lead's approval,
which is the `disposition_signoff` gate (ADR 0004, moved by ADR 0015). The build refuses a critic on one of its executors' models (K-12, ADR
0010), reading each agent's deployed model from its kubernetes/configmap.yaml. Each step's schemas are read from its agent's schemas/ folder,
so run this again after changing them. Standard library only.
The intake (s0, ADR 0008) checks the Discovery package in code before the planner starts, and fails the run if the
package is missing, changed, incomplete, from a failed critic, or from another run.
The report (s7, ADR 0015) is written by a model from facts worked out in code and carries the disposition_signoff
review point. Only after it is continued do the package (s8, ADR 0014: every step's output in one Assessment package,
in code, for Strategy and the later Features) and the approved report's page (s9) run.
Front steps (ADR 0002):
a1 Artifacts: the Discovery run's artifacts, by exact id -> a1.artifacts
p1 Parse Assessment Files over a1's records -> p1.parsed_content, p1.file_metadata
@@ -42,7 +46,7 @@ STAMP = "2026-11-02T12:00:00"
WF = ["engagement_id", "run_id", "discovery_execution_id"]
# The rate card is a run input, its JSON as text (ADR 0009): an uploaded `file` reaches every step, and the parse
# step that reads the Discovery package would read it instead.
RATE_CARD_STEPS = ("s1", "s5", "s6")
RATE_CARD_STEPS = ("s1", "s5", "s6", "s8")
CRITIC_STATIC = {"critic_model": "gpt-5.4", "executor_models": ["gpt-5.6-sol"]} # ADR 0005
DOC = {"discovery_documents": "p1.parsed_content", "discovery_files": "p1.file_metadata"}
@@ -67,6 +71,37 @@ STEPS = [
"posture_status": "s4.status", "security_findings": "s4.findings",
"dispositions": "s5.dispositions", "signoff_queue": "s5.signoff_queue"}),
]
# The report (ADR 0015) is written by a model from facts worked out in code, and carries the disposition_signoff review
# point, so the Engagement Lead signs off from it. After that review point: the package for Strategy and the later
# Features (ADR 0014), every input copied unchanged (cp1-assessment-package-agent/app/package.py INPUTS names the part
# each one belongs to), and the approved report as a page.
REPORT_STEP, PACKAGE_STEP, PAGE_STEP = "s7", "s8", "s9"
REVIEW_STEP = REPORT_STEP
PACKAGE_INPUTS = {
"intake_verified": "s0.verified", "discovery_run_id": "s0.discovery_run_id",
"discovery_package_version": "s0.package_version", "discovery_critic_verdict": "s0.critic_verdict",
"discovery_provider": "s0.provider", "discovery_parts_verified": "s0.parts_verified",
"discovery_absent_parts": "s0.absent_parts", "discovery_source_gaps": "s0.source_gaps",
"discovery_agent_versions": "s0.agent_versions",
"plan_status": "s1.status", "preconditions": "s1.preconditions", "rate_card_id": "s1.rate_card_id",
"in_cloud_estate": "s1.in_cloud_estate", "application_source": "s1.application_source",
"applications": "s1.applications", "tasks": "s1.tasks", "coverage_gaps": "s1.coverage_gaps",
"plan_summary": "s1.summary", "plan_note": "s1.findings_note",
"profiles": "s2.profiles", "profile_gaps": "s2.gaps", "profile_note": "s2.findings_note",
"scores": "s3.scores", "high_risk_tail": "s3.high_risk_tail", "risk_note": "s3.findings_note",
"posture_status": "s4.status", "security_provider": "s4.provider", "posture_score": "s4.posture_score",
"controls": "s4.controls", "security_findings": "s4.findings", "pillars": "s4.pillars",
"injection_flags": "s4.injection_flags", "security_note": "s4.findings_note",
"dispositions": "s5.dispositions", "producer_signoff_queue": "s5.signoff_queue", "scoring_note": "s5.scoring_note",
"verdict": "s6.verdict", "checks": "s6.checks", "rejected_dispositions": "s6.rejected_dispositions",
"signoff_queue": "s6.signoff_queue", "unverified_claims": "s6.unverified_claims",
"critic_model": "s6.critic_model", "executor_models": "s6.executor_models", "critic_summary": "s6.summary",
}
# The report also reads the Discovery package, to say what each application is (ADR 0015 addendum).
REPORT_INPUTS = {**PACKAGE_INPUTS, "discovery_documents": "p1.parsed_content"}
PACKAGE_INPUTS = {**PACKAGE_INPUTS, "report_markdown": f"{REPORT_STEP}.report_markdown"}
STEPS += [(REPORT_STEP, "cp1-assessment-report-agent", 7, REPORT_INPUTS),
(PACKAGE_STEP, "cp1-assessment-package-agent", 8, PACKAGE_INPUTS)]
FRONT_OUTPUTS = {"a1": {"artifacts"},
"p1": {"parsed_content", "file_metadata", "total_files", "total_content_length"}}
PARSE_OUT = {"type": "object", "properties": {
@@ -143,8 +178,14 @@ def build(artifact_ids=()) -> dict:
continue
src_out = json.loads((ROOT / by_id[src_step] / "schemas" / "output.json").read_text())
assert src_field in src_out["properties"], f"{sid}: {source} is not an output of {by_id[src_step]}"
inputs = {"from_workflow": WF + (["discovery_approval_ref"] if sid == "s1" else [])
inputs = {"from_workflow": WF + (["discovery_approval_ref"] if sid in ("s1", REPORT_STEP, PACKAGE_STEP) else [])
+ (["artifact_ids"] if sid == PACKAGE_STEP and not artifact_ids else [])
+ (["rate_card_document"] if sid in RATE_CARD_STEPS else []), "from_steps": from_steps}
if sid in (REPORT_STEP, PACKAGE_STEP):
# Which agent versions produced the run, as built; for the package, a fixed Discovery artifact id too.
inputs["static"] = {"agent_versions": {f: field(ROOT / f / "agent.yaml", "version")
for _, f, _, _ in STEPS if f != folder},
**({"artifact_ids": list(artifact_ids)} if artifact_ids and sid == PACKAGE_STEP else {})}
if sid == "s6":
for k in CRITIC_STATIC:
assert k in input_schema["properties"], k
@@ -155,13 +196,14 @@ def build(artifact_ids=()) -> dict:
depends = sorted({s.split(".")[0] for s in from_steps.values()} | ({"s0"} if sid == "s1" else set()))
connections += [{"source": d, "target": sid} for d in depends]
hitl = {"enabled": False}
if sid == "s6":
if sid == REVIEW_STEP:
hitl = {"enabled": True, "checkpoint_type": "approval",
"review_focus": "disposition_signoff: the Engagement Lead reads the critic's verdict and every "
"disposition in signoff_queue (confidence below 0.80, and every repurchase or "
"relocate) before Strategy starts. A fail verdict is never approved.",
"review_focus": "disposition_signoff: the Engagement Lead reads the Assessment report: the critic's "
"verdict and every disposition in the sign-off queue (confidence below 0.80, and "
"every repurchase or relocate). Continuing signs off that queue (ADR 0014) before "
"Strategy starts. A fail verdict is never approved.",
"timeout_hours": 72, "allowed_actions": ["continue", "rerun", "abort"], "max_reruns": 2}
y = {"s2": -160, "s3": 0, "s4": 160}.get(sid, 0)
y = {"s2": -160, "s3": 0, "s4": 160, PACKAGE_STEP: 160}.get(sid, 0)
steps.append({
"id": sid, "agent": agent_id, "agent_type": agent_id, "name": f"{display} Step",
"timeout_minutes": AGENT_STEP_TIMEOUT_MINUTES, "type": "sequential", "executionLevel": group,
@@ -175,6 +217,16 @@ def build(artifact_ids=()) -> dict:
"agent_metadata": {"agent_name": agent_id},
"inputSchema": input_schema, "outputSchema": output_schema}})
agents.append({"name": agent_id, "display_name": display, "source_url": url, "category": "agent"})
# After the review point: the approved report as a page, as this run's own artifact (as Discovery's, F01 ADR 0012).
page = _activity_step(PAGE_STEP, "format_for_ui", "Format Assessment report for UI",
"The approved Assessment report, for display as a page.",
{"from_steps": {"input_data": f"{REPORT_STEP}.final_html"}, "static": {"format_type": "html"}},
{"type": "object", "properties": {"input_data": {"type": "string"}, "format_type": {"type": "string"}}},
{"type": "object", "properties": {"data": {"type": "object"}}}, y=-160)
page.update({"depends_on": [REPORT_STEP], "executionLevel": 8, "parallelGroup": 8,
"position": {"x": 320 * 7, "y": -160}})
steps.append(page)
connections.append({"source": REPORT_STEP, "target": PAGE_STEP})
description = (f"{NAME}: the Discovery run's outputs and the rate card in; then "
+ "; then ".join(field(ROOT / f / "agent.yaml", "display_name") for _, f, _, _ in STEPS) + ".")
tags = ["cp1", "public-cloud", "agentic-landing-zone"]