Content
75%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A highly actionable, well-structured investigation skill: exact pipeline IDs, tool invocations, and error signatures leave no ambiguity about what to run, and the router design points to a canonical facts doc instead of restating it. The main weakness is token efficiency — the Jun 2026 root-cause narrative and volatile branch-train state are inlined time-sensitive detail that would live better in a reference or old-patterns section.
Suggestions
Move the 'Root cause of the Jun 2026 inflight/current outage' postmortem into a reference file (e.g. references/known-outages.md) or an explicitly marked historical section, keeping only the durable triage lesson inline: confirm IsPackable evaluates true for the workload packs in the pack.binlog and that the pack step's Build (hence _GenerateVSWorkloadProps) actually ran.
Move the volatile branch-train state (the .NET 11 train, net11 / candidate/11.0.1xx-sr1 / inflight/11.0.1xx-sr2 lane names) into the canonical maui-ci-facts.md or a clearly labeled 'current state' section, keeping only the Get-ReleaseRolloverPlan.ps1 resolution command in SKILL.md so the body does not rot as trains roll over.
Tighten the deployment-evidence paragraph ('Do not treat the local proposal, public branch existence...') into a one-line checkpoint rule — 'resolve the nightlyRef from deployed state and record its commit; if deployment evidence is missing, report the gap' — to cut prose while preserving the guard against querying a convenient green build.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Most of the body is dense, non-obvious operational fact (pipeline IDs, internal org/project, the MSB4019 signature), but the inlined "Root cause of the Jun 2026 inflight/current outage" postmortem is a ~180-word incident narrative carrying dates, PR numbers, and a commit hash (543b1ebeb7), and the branch-train paragraph pins volatile state ("the configured .NET 11 train", "candidate/11.0.1xx-sr1") — time-sensitive details not placed in an old-patterns/deprecated section, which the guidelines say to penalize. This fits anchor 3 ("mostly efficient but includes some... explanation or could be tightened"); not 4 because the postmortem is more than a minor trim, not 2 because there is no padding or explanation of concepts Claude already knows. | 3 / 5 |
Actionability | The body is fully executable: a copy-paste PowerShell command ("& ./eng/scripts/Get-ReleaseRolloverPlan.ps1 -ResolveStateOnly | ConvertFrom-Json"), a fully parameterized investigation chain ("azdo_builds with definitionId: 1095, branch: <resolved nightlyRef>" → "azdo_search_timeline (filter failed)" → "azdo_search_log"), an exact error signature ("workloads.csproj(30,3): error MSB4019"), exact pipeline definition IDs (302/314/313/1095), and a named fix ("./eng/ingest-maven-deps.sh") — matching anchor 5 (copy-paste ready, covers the common cases). | 5 / 5 |
Workflow Clarity | "Using this skill" gives a clear numbered 1–4 sequence (invoke ci-analysis → apply MAUI corrections → baseline comparison + merge-readiness criteria → escalate to helix-investigation), and the nightly path embeds explicit checkpoints ("confirm whether it's a one-off or a multi-day streak", "record its commit as evidence", "If deployment evidence is missing, report that gap"). This matches anchor 4 (clear sequence, most checkpoints present); not 5 because the checkpoints sit in prose rather than structured validate-and-branch steps within the sequence itself. No destructive/batch cap applies since the workflow is read-only investigation. | 4 / 5 |
Progressive Disclosure | The body is a router-style overview with a clearly signaled, one-level-deep canonical reference (".github/docs/maui-ci-facts.md — read it", its contents enumerated bullet-by-bullet) plus well-marked companion/escalation skills (ci-analysis, helix-investigation, review-test-failures). This matches anchor 4 (good structure, references mostly clear, minor organization gaps); not 5 because the ~48-line nightly section, including the incident postmortem, is inlined where a split reference file would fit — no bundle files exist to absorb it. Not 3: the inline content is bounded and headers make navigation easy. | 4 / 5 |
Total | 16 / 20 Passed |