From silver-bullet
Classifies change requests into three tiers (trivial, medium, complex) and routes to the appropriate workflow: direct edit, quality-gates pipeline, or feature escalation.
How this skill is triggered — by the user, by Claude, or both
Slash command
/silver-bullet:silver-fast <description of change><description of change>The summary Claude sees in its skill listing — used to decide when to auto-load this skill
SB triage spec for parent orchestrator. **Tier 1** still routes to a FAST worker (not parent edits). Tier 2+ escalates to fuller composer queues.
SB triage spec for parent orchestrator. Tier 1 still routes to a FAST worker (not parent edits). Tier 2+ escalates to fuller composer queues.
| Tier | Criteria | Routes to |
|---|---|---|
| Tier 1 (Trivial) | ≤3 files AND no logic changes AND no src//app code paths | parent spawns FAST worker |
| Tier 2 (Medium) | 4-10 files OR logic change in ≤3 files OR dependency update | silver:quality-gates (pre-plan), silver:plan, silver:validate, silver:execute, silver:verify; optional silver:context / silver:research when signals match |
| Tier 3 (Complex) | >10 files OR cross-cutting OR schema change OR new capability | silver:feature |
Note: Tier 2+ fast-path work starts a lightweight workflow tracker (
scripts/workflows.sh start /silver:fast ...) soworkflow-chain-guardand delivery gates can observe the path. Tier 1 remains direct edit with verification only.
Display banner:
SILVER BULLET ► FAST PATH
Change: {$ARGUMENTS or "(not specified)"}
Analyze $ARGUMENTS to classify into one of three tiers. Classification is autonomous — no interactive user prompt.
Tier 1 (Trivial):
src_pattern paths (typically src/, hooks/, scripts/, lib/, app/).py, .ts, .js, .go, .rs, .sh, .rb, .java, .kt, .swift, .tf, .hcl logic changeTier 2 (Medium):
Tier 3 (Complex):
10 files OR cross-cutting concern OR schema change OR new capability
Ambiguity rules (always bias toward the safer/more thorough tier):
Display classification:
Classification: Tier {N} ({Trivial|Medium|Complex})
Routing to: {direct SB edit|SB lifecycle slice|silver:feature}
Only reached when Step 0 classifies as Tier 1 (Trivial).
Make the small edit directly in the current session. Keep the change to the classified trivial scope and run the smallest relevant verification.
After the trivial edit and verification complete, run scope expansion check (Step 4).
Tier 1 discipline: Do not misclassify logic changes as trivial to bypass workflow tracking. The ${SB_RUNTIME_HOME_ROOT}/.silver-bullet/trivial marker only applies to genuine typo/config-only sessions; any src logic change requires Tier 2+ or silver:feature.
Trivial bypass mid-session: SessionStart clears any stale ${SB_RUNTIME_HOME_ROOT}/.silver-bullet/trivial marker; the first Write/Edit removes it if recreated (PostToolUse hook). If you start Tier 1 then expand scope into logic changes, workflow-chain-guard and delivery gates apply normally. Re-classify as Tier 2+ and start scripts/workflows.sh start /silver:fast before further implementation edits.
If scope remained ≤3 files, display completion banner:
SILVER BULLET ► FAST PATH COMPLETE
Change: {$ARGUMENTS}
Files modified: {count} (confirmed ≤3)
Status: verified
Only reached when Step 0 classifies as Tier 2 (Medium).
Before invoking SB lifecycle skills, detect which gates to apply by scanning $ARGUMENTS for signals:
Signal detection:
| Gate | Signal words in $ARGUMENTS |
|---|---|
silver:context | "not sure", "unclear", "multiple approaches", "options", "decide", "which", "should we", "trade-off", "either...or" |
silver:research | "new library", "unfamiliar", "investigate", "evaluate", "compare", "never used", "first time", "unknown", "explore options" |
Gate composition rules:
silver:plan, silver:validate, silver:execute, and silver:verify for Tier 2.silver:quality-gates (pre-plan design-time) before planning when starting the Tier 2 tracker — workflow-chain-guard enforces this marker..planning/SPEC.md is absent, silver:validate runs in plan-only mode (PLAN.md completeness — see skills/silver-validate/SKILL.md Step 0). Do not run /silver:spec unless product requirements need a formal SPEC.silver:context or silver:research only when triggered by the signal table above.workflow-chain-guard for Tier 2 requires silver-quality-gates (design marker), silver-plan, and silver-validate before implementation edits — not silver-context.Display detected signals:
Detected signals:
Ambiguity: {yes/no} {reason if yes}
Novel domain: {yes/no} {reason if yes}
Gates: {silver:context silver:research | (none)} + silver:plan silver:validate silver:execute silver:verify (always)
Invoke the selected SB lifecycle slice with $ARGUMENTS.
Workflow tracker (Tier 2 — mandatory): Before the first implementation edit, start the fast-path workflow tracker:
if [[ -x scripts/workflows.sh ]]; then
SB_WORKFLOWS_BIN="scripts/workflows.sh"
else
SB_WORKFLOWS_BIN="$(
for root in \
"${SB_RUNTIME_HOME_ROOT}/plugins/cache/alo-labs/silver-bullet/current" \
"$HOME/.codex/plugins/cache/alo-labs-codex/silver-bullet/current"; do
if [[ -x "$root/scripts/workflows.sh" ]]; then
printf "%s\n" "$root/scripts/workflows.sh"
break
fi
done
)"
fi
if [[ -n "${SB_WORKFLOWS_BIN:-}" ]]; then
SB_WORKFLOW_ID=$("$SB_WORKFLOWS_BIN" start /silver:fast "$ARGUMENTS" "QUALITY GATE,plan,validate,execute,verify")
export SB_WORKFLOW_ID
fi
After each invoked lifecycle skill completes, run "$SB_WORKFLOWS_BIN" complete-flow "$SB_WORKFLOW_ID" "<skill-name>".
After the SB lifecycle slice completes, run scope expansion check (Step 4).
The Tier 2 slice (quality-gates → plan → validate → execute → verify) intentionally stops at
verification. It does not record the post-execution deploy chain
(silver:review-request, silver:review, silver:review-triage, silver:secure,
silver:validate, silver:branch-finish, silver:ship).
Consequences and routing:
silver:verify — the Stop gate only requires the
planning floor, so this completes cleanly.required_deploy list. A bare Tier 2 slice will be blocked
at gh pr create / gh release create / deploy because the review, security,
validate, branch-finish, and ship markers are missing. Before raising a PR you
MUST either:
silver:feature) —
silver:review-request → silver:review → silver:review-triage → silver:verify → security → silver:secure → silver:validate → silver:quality-gates → silver:branch-finish → silver:completion-audit → silver:ship — then deliver, orsilver:feature (Step 3), which owns the full lifecycle and
deploy chain end-to-end.Do not attempt to bypass the deploy gate by editing state files or skipping skills.
After Tier 2 execution, any item scoped out during execution MUST be filed via /silver:add:
Skill(skill="silver:add", args="<description of deferred item>")
Note: Tier 1 (trivial changes) → no capture needed. Tier 3 → escalates to /silver:feature, which handles its own deferred-item capture.
Only reached when Step 0 classifies as Tier 3 (Complex).
Display:
Change exceeds fast-path complexity. Routing to silver:feature.
Reason: {specific reason — e.g., "touches >10 files", "cross-cutting concern", "schema change", "new capability"}
Invoke silver:feature through the active runtime's SB-recognized skill invocation channel with $ARGUMENTS. Exit silver:fast.
After Tier 1 or Tier 2 execution completes, check if scope expanded beyond the current tier.
During Tier 1: If files modified > 3:
During Tier 2: If files modified > 10:
Escalation is autonomous — no interactive user prompt needed. Display escalation banner:
FAST PATH ESCALATION
Reason: Scope expanded from {original tier} to {new tier}
Files affected: {count}
Routing to: {SB lifecycle slice|silver:feature}
Then invoke the target workflow. On escalation to silver:feature, pass the updated scope description so /silver can classify appropriately.
npx claudepluginhub alo-exp/silver-bullet --plugin silver-bulletGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.