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
1---2name: risk-analysis-33description: >-4 Probability × impact risk assessment for curation and planning decisions. Categorize risks5 (quality, coherence, scope creep, token budget, breaking change, staleness), score by likelihood6 × impact, and route by risk score to ship / gate / ADR. Use before shipping a change, when7 accepting an external artifact, when running a pre-release gate, or when triaging findings.8metadata:9 source: https://skillmd.com/skills/fabioc-aloha/risk-analysis-310---1112# Curation Risk Analysis1314> **ACT Tenet VI**: Match rigor to stakes. Reversible decisions deserve speed; irreversible ones deserve doubt.1516Applied to curation work: accepting a bad skill is reversible (revert next patch). Shipping a broken release to consumers is expensive to undo.1718## Shared Knowledge Gate1920Before evaluating non-trivial, repeated, or consequential risk, consult21[`scout-knowledge-base`](../scout-knowledge-base/SKILL.md) for relevant prior22decisions, failure modes, procedures, and gotchas.23241. Search the shared index using the risk's problem-specific terms.252. Read only records whose signals or preconditions match the current risk.263. Treat records as evidence and context, not authority; verify whether their27 conditions still hold.284. If the knowledge base is unavailable, state that once and continue with the29 current evidence. Do not create a local fallback while assessing risk.3031Skip this gate for trivial, isolated work. If shared storage needs first-time32configuration, route to `scout-shared-data-setup` after the current task or33when the user asks to enable it.3435## Risk Categories (Curation-Specific)3637| Category | Examples |38|----------|----------|39| **Quality** | Skill has bugs, stale references, or misleading content |40| **Coherence** | New content conflicts with existing brain architecture |41| **Scope creep** | Feature belongs in a different repo or surface (not Core / Mall / Steward) |42| **Token budget** | Addition pushes Core past 25K instruction budget |43| **Breaking change** | Rename/remove that breaks heir upgrades |44| **Staleness** | Store/skill stays past its useful life, misleading by presence |4546## Risk Matrix for Curation Decisions4748| | Low Impact | Medium Impact | High Impact |49|---|-----------|---------------|-------------|50| **High Likelihood** | Ship with note | Gate on review | Full ACT pass + ADR |51| **Medium Likelihood** | Ship | Ship with note | Gate on review |52| **Low Likelihood** | Ship | Ship | Ship with note |5354### Impact Calibration for This Repo5556| Level | Curation Impact |57|-------|-----------------|58| **High** | Breaks heir upgrades, violates Cardinal Rule, token budget breach |59| **Medium** | Quality gap ships to heirs, coherence drift, stale reference |60| **Low** | Cosmetic issue, suboptimal wording, minor inconsistency |6162### Likelihood Calibration6364| Level | Probability | Curation Signal |65|-------|-------------|-----------------|66| **High** | >60% | Multiple heirs affected, or same issue seen before |67| **Medium** | 20-60% | Plausible failure mode, no prior occurrence |68| **Low** | <20% | Theoretical risk, strong mitigations in place |6970## Decision Routing by Risk Score7172| Score (L×I) | Route |73|-------------|-------|74| Low × Low | Ship immediately, log in curation-log |75| Any Medium | Ship with trimmed ACT pass visible in curation-log |76| Any High | Full ACT pass, write ADR, require explicit approval |77| High × High | Surface as ADR + direct Fabio conversation before acting |7879## Risk Register (Running)8081Maintained 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:8283- Release candidates (per your project's release process)84- Adopting an external artifact that scored below your project's acceptance bar85- Structural changes to your project's brain architecture8687## Would Revise If8889Revise by **2026-11-30** or sooner if any of the following fires:9091- The risk matrix produces no differentiation (everything routes to "Ship with note") across 10+ curation decisions in a quarter92- A curation decision that scored Low×Low ships and produces fleet-breaking impact ≥1 time (impact calibration too lax)93- 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)94- The Likelihood/Impact tables drift from observed reality (most "High Likelihood" rows turn out Low) ≥3 times in a quarter