| name | risk-analysis |
| description | Probability × impact risk assessment for curation and planning decisions. Categorize risks (quality, coherence, scope creep, token budget, breaking change, staleness), score by likelihood × impact, and route by risk score to ship / gate / ADR. Use before shipping a change, when accepting an external artifact, when running a pre-release gate, or when triaging findings. |
Curation Risk Analysis
ACT Tenet VI: Match rigor to stakes. Reversible decisions deserve speed; irreversible ones deserve doubt.
Applied to curation work: accepting a bad skill is reversible (revert next patch). Shipping a broken release to consumers is expensive to undo.
Shared Knowledge Gate
Before evaluating non-trivial, repeated, or consequential risk, consult
scout-knowledge-base for relevant prior
decisions, failure modes, procedures, and gotchas.
- Search the shared index using the risk's problem-specific terms.
- Read only records whose signals or preconditions match the current risk.
- Treat records as evidence and context, not authority; verify whether their
conditions still hold.
- If the knowledge base is unavailable, state that once and continue with the
current evidence. Do not create a local fallback while assessing risk.
Skip this gate for trivial, isolated work. If shared storage needs first-time
configuration, route to scout-shared-data-setup after the current task or
when the user asks to enable it.
Risk Categories (Curation-Specific)
| Category | Examples |
|---|
| Quality | Skill has bugs, stale references, or misleading content |
| Coherence | New content conflicts with existing brain architecture |
| Scope creep | Feature belongs in a different repo or surface (not Core / Mall / Steward) |
| Token budget | Addition pushes Core past 25K instruction budget |
| Breaking change | Rename/remove that breaks heir upgrades |
| Staleness | Store/skill stays past its useful life, misleading by presence |
Risk Matrix for Curation Decisions
| Low Impact | Medium Impact | High Impact |
|---|
| High Likelihood | Ship with note | Gate on review | Full ACT pass + ADR |
| Medium Likelihood | Ship | Ship with note | Gate on review |
| Low Likelihood | Ship | Ship | Ship with note |
Impact Calibration for This Repo
| Level | Curation Impact |
|---|
| High | Breaks heir upgrades, violates Cardinal Rule, token budget breach |
| Medium | Quality gap ships to heirs, coherence drift, stale reference |
| Low | Cosmetic issue, suboptimal wording, minor inconsistency |
Likelihood Calibration
| Level | Probability | Curation Signal |
|---|
| High | >60% | Multiple heirs affected, or same issue seen before |
| Medium | 20-60% | Plausible failure mode, no prior occurrence |
| Low | <20% | Theoretical risk, strong mitigations in place |
Decision Routing by Risk Score
| Score (L×I) | Route |
|---|
| Low × Low | Ship immediately, log in curation-log |
| Any Medium | Ship with trimmed ACT pass visible in curation-log |
| Any High | Full ACT pass, write ADR, require explicit approval |
| High × High | Surface as ADR + direct Fabio conversation before acting |
Risk Register (Running)
Maintained in your project's audit trail (Alex ACT itself uses operations/ledgers/brain-qa-changelog.md) — every entry carries implicit risk assessment via the Severity column. Explicit risk registers warranted only for:
- Release candidates (per your project's release process)
- Adopting an external artifact that scored below your project's acceptance bar
- Structural changes to your project's brain architecture
Would Revise If
Revise by 2026-11-30 or sooner if any of the following fires:
- The risk matrix produces no differentiation (everything routes to "Ship with note") across 10+ curation decisions in a quarter
- A curation decision that scored Low×Low ships and produces fleet-breaking impact ≥1 time (impact calibration too lax)
- A High×High decision is over-routed to ADR when a trimmed pass would have sufficed ≥3 times in a quarter (rigor calibration too tight)
- The Likelihood/Impact tables drift from observed reality (most "High Likelihood" rows turn out Low) ≥3 times in a quarter