Skip to content

thedotmack/claude-mem / plugin/skills/agent-cost-report

agent-cost-report

Verified

Believable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json.

FileKindSize
.gitignoreother100 B
CHECKSUMS.txtmarkdown2.1 KB
SKILL.mdskill md21.3 KB
VERIFICATION.mdmarkdown12.3 KB
scripts/acr.pyexecutablescript12.2 KB
scripts/acr/__init__.pyscript105 B
scripts/acr/behavior.pyscript17.5 KB
scripts/acr/classify.pyscript5.5 KB
scripts/acr/costs.pyscript6.3 KB
scripts/acr/devices.pyscript5.4 KB
scripts/acr/evidence.pyscript9.2 KB
scripts/acr/labels.pyscript7.6 KB
scripts/acr/measure.pyscript5.6 KB
scripts/acr/mistakes.pyscript23.1 KB
scripts/acr/owner.pyscript2.9 KB
scripts/acr/patterns.pyscript26.0 KB
scripts/acr/period.pyscript3.8 KB
scripts/acr/prices.pyscript5.0 KB
scripts/acr/render.pyscript52.7 KB
scripts/acr/rollup.pyscript31.7 KB
scripts/acr/rules.jsonconfig3.5 KB
scripts/acr/rules.pyscript4.5 KB
scripts/acr/sample.pyscript3.6 KB
scripts/acr/snapshot.pyscript2.0 KB
scripts/acr/sync.pyscript3.7 KB
scripts/acr/transcripts.pyscript7.4 KB
scripts/acr/wins.pyscript21.0 KB
scripts/tests/_db.pyscript3.6 KB
scripts/tests/_paths.pyscript337 B
scripts/tests/fixtures/codex/sessions/2026/09/18/rollout-codex-1.jsonlother664 B
scripts/tests/fixtures/make_report_fixtures.pyscript2.0 KB
scripts/tests/fixtures/prices_sample.jsonconfig611 B
scripts/tests/fixtures/projects/-workspace-proj-a/sess-aaaa.jsonlother2.9 KB
scripts/tests/fixtures/projects/-workspace-proj-b/sess-bbbb.jsonlother537 B
scripts/tests/fixtures/report-1day.jsonconfig20.1 KB
scripts/tests/fixtures/report-30day.jsonconfig48.1 KB
scripts/tests/fixtures/report-7day.jsonconfig29.1 KB
scripts/tests/fixtures/report-empty.jsonconfig13.8 KB
scripts/tests/fixtures/report-session.jsonconfig16.6 KB
scripts/tests/test_behavior.pyscript29.6 KB
scripts/tests/test_devices.pyscript5.9 KB
scripts/tests/test_labels.pyscript8.6 KB
scripts/tests/test_measure.pyscript6.8 KB
scripts/tests/test_period.pyscript7.0 KB
scripts/tests/test_prices.pyscript5.3 KB
scripts/tests/test_render.pyscript12.1 KB
scripts/tests/test_render_encoding.pyscript1.5 KB
scripts/tests/test_rollup.pyscript19.3 KB
scripts/tests/test_snapshot.pyscript2.8 KB
scripts/tests/test_sync.pyscript2.5 KB
scripts/tests/test_transcript_encoding.pyscript2.7 KB
scripts/tests/test_transcripts.pyscript8.1 KB
scripts/tests/test_wins.pyscript15.7 KB
verification/precision-2026-09-26.jsonconfig308 B

scripts/acr/sample.py

View on GitHub
"""Phase 8.5: `behavior-sample` writes a Markdown spot-check sheet from <run>/behavior.json and report.json:
N random human-tagged and N bot-tagged user turns (R2 check), N flagged turns or episodes per summary-tile
pattern, and 10 random unflagged sessions. Each line cites the session and time so a reviewer can mark it
true / false / unsure. Precision per pattern goes back into `rollup --precision <json>` (tile gate, 70%).
"""
import collections
import json
import os
import random

from . import patterns

TILE_GROUPS = [("P1_invented_gates", ["P1_invented_gates"]), ("P6+P7 made it up / false done", ["P6_fake_output", "P7_false_done"]),
               ("P2_broke_things", ["P2_broke_things"]), ("P3_wrong_model", ["P3_wrong_model"])]


def write(report_path, out_path, per_metric=20, seed=7):
    with open(report_path, encoding="utf-8") as fh: d = json.load(fh)
    bp = os.path.join(os.path.dirname(os.path.abspath(report_path)), "behavior.json")
    with open(bp, encoding="utf-8") as fh: b = json.load(fh)
    rnd = random.Random(seed); L = [f"# Behavior spot-check · {d['window'].get('start_pt')} → {d['window'].get('end_exclusive_pt')} · seed {seed}", "",
                                    "Mark each line `true`, `false` or `unsure` in the first column. Excerpts are scrubbed and cut to 160 chars.", ""]
    turns = b.get("user_turns", [])
    for author in ("human", "bot"):
        pool = [t for t in turns if t["author"] == author]; pick = rnd.sample(pool, min(per_metric, len(pool)))
        L += [f"## User turns tagged {author} ({len(pick)} of {len(pool)})", "", "| verdict | session | time | source | excerpt |", "|---|---|---|---|---|"]
        L += [f"| | `{t['session'][:12]}` | {t['ts_pt']} | {t['source']} | {t['excerpt'].replace('|', '/')} |" for t in pick]
        L.append("")
    flags = [f for f in b.get("flags", []) if not f.get("unconfirmed")]; eps = b.get("episodes", [])
    for name, keys in TILE_GROUPS:
        pool = [dict(kind="turn", **f) for f in flags if f["pattern"] in keys] + [dict(kind="episode", **e) for e in eps if e["pattern"] in keys]
        pick = rnd.sample(pool, min(per_metric, len(pool)))
        L += [f"## {name} ({len(pick)} of {len(pool)} flagged)", "", "| verdict | kind | session | time | basis | excerpt |", "|---|---|---|---|---|---|"]
        for x in pick:
            sid = x.get("session") or ""; ts = x.get("ts_pt") or ""
            L.append(f"| | {x['kind']} | `{sid[:12]}` | {ts} | {(x.get('basis') or x.get('cost_status') or '').replace('|', '/')} | {(x.get('excerpt') or '').replace('|', '/')} |")
        L.append("")
    flagged_sessions = {f["session"] for f in flags} | {e["session"] for e in eps}
    unflagged = [s for s in b.get("sessions", []) if s["session"] not in flagged_sessions]
    pick = rnd.sample(unflagged, min(10, len(unflagged)))
    L += [f"## Unflagged sessions ({len(pick)} of {len(unflagged)}) — look for obvious misses", "", "| missed? | session | tag | turns | tool errors | last message excerpt |", "|---|---|---|---|---|---|"]
    L += [f"| | `{s['session'][:12]}` | {s['tag']} | {s['turns']} | {s['tool_errors']} | {(s.get('last_text') or '').replace('|', '/')} |" for s in pick]
    L += ["", "## Precision (fill in)", "", "| pattern | true | false | unsure | precision |", "|---|---|---|---|---|"] + [f"| {name} | | | | |" for name, _ in TILE_GROUPS] + [""]
    with open(out_path, "w", encoding="utf-8") as fh: fh.write("\n".join(L))
    return out_path, dict(human=sum(1 for t in turns if t["author"] == "human"), bot=sum(1 for t in turns if t["author"] == "bot"), flags=len(flags), episodes=len(eps), unflagged=len(unflagged))
deprecation-and-migrationaddyosmani/agent-skillsManages deprecation and migration. Use when removing old systems, APIs, or features. Use when migrating users from one implementation to another. Use when migrating a database schema in production, such as renaming or dropping a column without downtime (expand/contract). Use when deciding whether to maintain or sunset existing code.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
performance-optimizationaddyosmani/agent-skillsOptimizes application performance across frontend, backend, queries, and databases. Use when performance requirements exist, when you suspect performance regressions, when Core Web Vitals or load times need improvement, when N+1 query patterns need fixing, or when profiling reveals bottlenecks.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
backup-setupthedotmack/claude-memSet up, check, or restore local claude-mem database snapshots. Use when the user says "set up backups", "back up my memories", "backup status", or "restore my database".Stars: 118kUpdated Claude CodeCodexVerifiedInstall