performance

A command that reports how an audience rated a recorded session and, when enough comparison data exists, how it differed from the project’s own baseline. The baseline is a collection of earlier results for similar tasks.

In plain words
What is it for?
Use it to check the required judging script and DuckDB, then view the session verdict and any earned baseline comparison.
Why use it?
It separates the general reaction to a session from evidence that it performed better or worse than previous comparable sessions.

Command

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add commands/chrisdbaldwin/masques/performance
Clone the repo
git clone --depth 1 https://github.com/ChrisDBaldwin/Masques
Per session 19 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,508 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00019 $0.01508
Opus 5 $0.00010 $0.00754
Sonnet 5 $0.00004 $0.00302
Haiku 4.5 $0.00002 $0.00151

Measured yesterday against content hash b3a5879faff4, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

performance scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

commands/performance.md · 159 lines

How it starts

The opening of the file, as written. The whole thing — 159 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Performance Command

Show how the always-on audience scored a session. Output has two layers (PRD D4):

  • Layer A — the house reaction (always, from session one): a single 7-point verdict — perfect · great · good · neutral · bad · awful · detracting. An honest read of how the session went, never a claim that the masque caused it.
  • Layer B — lift (only once earned): how the masque compares to your own baseline corpus on the same task-class — a delta, never a bare level. Hidden until the baseline is thick enough. The audience never shows a misleading number.

This replaces the old five-proxy "masque verdict." Those proxies were masque-agnostic — they scored the same with or without a masque — so they are demoted to supporting_signals (context), never the verdict (PRD D7).

Instructions

Step 1: Check Prerequisites

  1. Judge script present:
test -f ${CLAUDE_PLUGIN_ROOT}/services/judge/judge.sh && echo ok || echo missing
  1. DuckDB present:
command -v duckdb >/dev/null && echo ok || echo missing

If missing: "Scoring requires DuckDB. Install: brew install duckdb." 3. Telemetry data present:

test -f ${CLAUDE_PLUGIN_ROOT}/services/collector/data/logs.jsonl && echo ok || echo "no data"

If no data: the audience hasn't captured anything yet — "Run /audience seat to open the house (it stays open), then start a session." Note capture needs the Claude Code OTEL env (see /audience seat).

Step 2: Identify the session to score

By default the judge scores the most recent session. To score this session explicitly, pass its id (which is also what /don records for attribution):

echo "$CLAUDE_CODE_SESSION_ID"

Optionally read .claude/masque.session.yaml to mention the active masque by name in your framing.

Step 3: Run the judge

TARGET_SESSION="$CLAUDE_CODE_SESSION_ID" ${CLAUDE_PLUGIN_ROOT}/services/judge/judge.sh

(Omit TARGET_SESSION to score the most recent session.) Optional knobs:

  • BASELINE_MIN=N — baseline sessions per task-class before Layer-B lift shows (default 5).
  • RUBRIC_BAND=<band> — if a rubric judge (an LLM pass, or a Witness-masque agent — D7) has read the session against the masque's rubric and produced a band, pass it here; the judge will use it for Layer A (judge: rubric) instead of the activity fallback.

Read the full file on GitHub · 159 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 159 lines · 19 tokens per session scan A b3a5879faff4

Subscribe to this mod's changes

performance is a command published in the GitHub repository ChrisDBaldwin/Masques (14 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 19 tokens to every session and 1,508 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.