{"boundary_questions": [], "candidates": [{"boundary_confidence": "high", "candidate_id": "inference-arbitrage/offload-analyst/adhoc-json-reshaping", "determinism_tests": {"T1": true, "T2": true, "T3": true, "T4": true, "T5": true}, "digest_schema": null, "escalation_path": null, "falsifiability": {"examples": true, "overrule_case": true, "signature": true}, "issue": "kotkan/claude-plugin-inference-arbitrage#24", "measurement": {"cost_per_invocation": 32673, "invocations": 26, "offload_value": 849503, "offload_waste": 219850, "share_of_plugin": 0.4009}, "measurement_strength": "measured", "position": "llm-over-script-digest", "signature": ["Bash(command=, description=)", "Bash(command=, description=)", "Bash(command=, description=)", "Bash(command=, description=)"], "signature_hash": "4cd9a21f6252d067", "signature_source": "scan-ngram", "skill": "inference-arbitrage:offload-analyst", "verdict": "file"}], "coverage": {"attributed_turns": 263, "ratio": 0.226}, "measurements_by_signature": {"02c9e27c5d8685b0": {"cost_per_invocation": 23653, "invocations": 3, "offload_value": 70959, "share_of_plugin": 0.0335}, "4cd9a21f6252d067": {"cost_per_invocation": 27257, "invocations": 3, "offload_value": 81773, "share_of_plugin": 0.0386}, "4e0b710f98812973": {"cost_per_invocation": 15856, "invocations": 4, "offload_value": 63427, "share_of_plugin": 0.0299}, "bd3cb224f4a0389a": {"cost_per_invocation": 13441, "invocations": 3, "offload_value": 40323, "share_of_plugin": 0.019}, "be7ae2818a8ec378": {"cost_per_invocation": 22715, "invocations": 3, "offload_value": 68146, "share_of_plugin": 0.0322}}, "notes": ["inference-arbitrage/offload-audit/filing-dispatch-hop: no-offload — position 4 — correctly done by inference. Reported as a finding, not a question.", "self-audit: inference-arbitrage plugin audited against itself", "wiki-checked: empty (confirmed via cluster:gitea-agent wiki_read 404 on Data/inference-arbitrage/snapshots.jsonl before running stability-classify)", "filing-dispatch-hop candidate graded no-offload (position pure-inference): filing-plan already does the mechanical marker/body construction as script; remaining Agent-dispatch hop is a deliberate credential-boundary architecture choice, not an unscripted algorithm; also thin evidence (share 0.52%, well under the 2% threshold)", "hypothesis 2 from the audit brief (marker-search fragility) verified false by reading bin/filing-plan directly: marker()/marker_re()/find_existing() already implement a tolerant-whitespace, exact-id regex match as a deterministic script step, not an inference judgment — no candidate, not graded", "hypothesis 3 from the audit brief (scan itself burning inference) verified false: bin/offload-scan and bin/plugin-inventory are pure Python, no LLM calls; the 22.6% coverage ratio this window is dilution from narrow window/session mix, not a cost problem in the scan tools"], "rubric_version": "1.0.0", "run_id": "2026-07-30T14:49:20Z", "target": {"name": "inference-arbitrage", "repo": "kotkan/claude-plugin-inference-arbitrage", "skills": ["boundary-rubric", "offload-analyst", "offload-audit", "offload-trend"], "source": "path", "version": "0.7.0"}, "tool_versions": {"cc-tokens": "/home/oleks/projects/claude-plugins/token-budget/bin/cc-tokens", "inference-arbitrage": "0.7.0"}, "totals": {"invocation_caveat": "invocation counts are a lower bound: two same-skill calls within 10 minutes of each other on the wall clock are merged into one reported invocation", "invocations": 52, "share_mechanical": 0.171, "weighted_tokens": 2119174}, "window": {"requested_since": "2026-06-30T14:39:24.064314+00:00", "requested_until": null, "sessions": 3, "since": "2026-07-29T18:42:20.877000+00:00", "until": "2026-07-30T14:40:05.115000+00:00"}}
1
Data/inference-arbitrage/snapshots.jsonl
oleks edited this page 2026-07-30 17:50:28 +03:00