Wraps
getTokenOptimizer().getCompactContext()
from
for cost-analysis queries. The bridge dynamically imports
with graceful fallback: when the package isn't installed,
is
and the skill exits cleanly. No MCP tool wraps
today (ADR-0002 §"Riskiest assumption"); we shell a Node one-liner instead.
-
Take the query — the single argument.
-
Invoke — run from anywhere under
so
resolves:
bash
( cd v3 && node ../plugins/ruflo-cost-tracker/scripts/compact.mjs "<QUERY>" )
The script imports
@claude-flow/integration/token-optimizer
(canonical export —
not , which would double the
extension via Node's
exports rule), calls
, and prints a markdown summary plus a JSON line via
.
-
Report — markdown table with: memories retrieved, tokens saved (bridge-reported), agentic-flow availability, cache hit rate. The script also emits a "bridge-reported, not measured against a no-RAG baseline" disclaimer. On bridge-unavailable: prints "agentic-flow not installed — bridge returns inert results." and exits cleanly.
CLAUDE.md root claims
ReasoningBank retrieval: -32%
tokens. The bridge's
is
query_tokens − compact_prompt_tokens
(token-optimizer.ts:141–143) — a heuristic,
not a baseline-measured saving. token-optimizer.ts:9–10 itself says:
"No fabricated metrics are reported — all stats reflect real measurements". This skill carries that disclaimer forward.
Booster-specific availability is
not exposed as a getter — observable only through
returning
. The canonical Tier 1 signal is
[AGENT_BOOSTER_AVAILABLE]
(see
).
not installed →
returns
{tokensSaved: 0, memories: []}
(line 116–124),
returns
,
falls back to anti-drift defaults. Skill exits cleanly with the "not available" message.
ADR-0002 Decision #2 + §"Riskiest assumption" ·
(singleton export) ·
docs/benchmarks/0002-baseline.md
(verification findings).