compare-tasks

Instructions for comparing two Harbor benchmark runs on the same task, where a run is a recorded set of benchmark trials.

In plain words
What is it for?
They locate matching trial folders, verify their task names, compare headline results, and inspect the recorded outputs to identify the cause of the difference.
Why use it?
They help explain why two runs produced different results instead of only reporting that their scores differ.

Skill for Claude CodeCodex

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 skills/aaif-goose/goose/compare-tasks
Any agent
npx skills add aaif-goose/goose --skill compare-tasks
Clone the repo
git clone --depth 1 https://github.com/aaif-goose/goose

Made for: Claude Code, Codex.

Per session 16 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,430 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 $0.00016 $0.02430
Opus 5 $0.00008 $0.01215
Sonnet 5 $0.00003 $0.00486
Haiku 4.5 $0.00002 $0.00243

Measured yesterday against content hash 6972b34c696f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

compare-tasks 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.

evals/harbor/.agents/skills/compare-tasks/SKILL.md · 228 lines

How it starts

The opening of the file, as written. The whole thing — 228 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Compare two harbor runs on one task

Use when given two harbor run names and a task name, and the goal is to understand why the two runs differ on that task — not just that they differ.

Inputs

  • RUN_A: harbor run name (e.g. sonnet46-full)
  • RUN_B: harbor run name (e.g. pi-sonnet46-full)
  • TASK: bare task name (e.g. extract-elf, not terminal-bench/extract-elf)
  • RUNS_DIR: defaults to evals/harbor/runs/ relative to the repo root

Procedure

1. Find each run's trial directory for the task

Harbor 0.8 names trial dirs <task>__<random-suffix> (e.g. extract-elf__bU3GHs4), not <task>.1. The suffix is unique per trial, so don't guess it — discover it from disk:

TRIAL_A_DIR=$(ls -d "$RUNS_DIR/$RUN_A/${TASK}__"*/ 2>/dev/null | head -1)
TRIAL_B_DIR=$(ls -d "$RUNS_DIR/$RUN_B/${TASK}__"*/ 2>/dev/null | head -1)

If either is empty, that run didn't include this task — stop and say so. (ls "$RUNS_DIR/$RUN_A/" shows what's there.)

If you want to confirm the match, every result.json carries task_name and trial_name:

jq '{task_name, trial_name}' "$TRIAL_A_DIR/result.json"

2. Headline facts

The fastest path is to let cmd.py task do it for you — it already prints status, reward, duration, tokens, turns, cost, error class, and the tail of the verifier stdout:

./evals/harbor/cmd.py task "$RUN_A" "$TASK"
./evals/harbor/cmd.py task "$RUN_B" "$TASK"

Only drop to raw jq against result.json if you need a field cmd.py task doesn't print. The actual shape (harbor 0.8 TrialResult):

jq '{
  reward: (.verifier_result.rewards.reward
           // (.verifier_result.rewards | to_entries | .[0].value)
           // null),
  rewards_all: .verifier_result.rewards,
  duration_seconds: ((.finished_at | fromdateiso8601) - (.started_at | fromdateiso8601)),
  input_tokens: .agent_result.n_input_tokens,
  cache_tokens: .agent_result.n_cache_tokens,
  output_tokens: .agent_result.n_output_tokens,
  cost_usd: .agent_result.cost_usd,
  error_type: .exception_info.exception_type,
  error_message: (.exception_info.exception_message // "" | split("\n")[0])
}' "$TRIAL_A_DIR/result.json"

Read the full file on GitHub · 228 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 · 228 lines · 16 tokens per session scan A 6972b34c696f

Subscribe to this mod's changes

compare-tasks is a skill published in the GitHub repository aaif-goose/goose (53,738 stars, last pushed today), licensed Apache-2.0. It adds 16 tokens to every session and 2,430 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.

Related

Other skills, from other repositories

release-notes

Generate user-facing release notes for Intelligent Terminal. Use when asked to write release notes, changelog, what-is-new summary, or prepare a release. Compares git commits between releases, looks up PR-linked issues and community contributors, then outputs formatted notes with "Verbed + Impact + Scenario" style…

microsoft/intelligent-terminal · 73 tokens

peer-review

Structured manuscript/grant review with checklist-based evaluation. Use when writing formal peer reviews with specific criteria methodology assessment, statistical validity, reporting standards compliance (CONSORT/STROBE), and constructive feedback. Best for actual review writing, manuscript revision. For evaluating…

xintaofei/codeg · 71 tokens

scientific-critical-thinking

Evaluate scientific claims and evidence quality. Use for assessing experimental design validity, identifying biases and confounders, applying evidence grading frameworks (GRADE, Cochrane Risk of Bias), or teaching critical analysis. Best for understanding evidence quality, identifying flaws. For formal peer review…

xintaofei/codeg · 63 tokens

statistical-analysis

Guided statistical analysis for research data - test selection, assumption checking, effect sizes, power analysis, Bayesian alternatives, and APA-formatted reporting. Use whenever a user wants to compare groups, test a hypothesis, analyze experimental or survey data, check statistical assumptions, compute required…

xintaofei/codeg · 111 tokens

statistical-power

Sample-size and statistical power calculations for planning studies. Use whenever someone asks "how many subjects/samples/replicates do I need", wants an a priori power analysis, a minimum detectable effect (MDE), a power curve, or needs to justify a sample size for a grant, IRB protocol, or pre-registration. Covers…

xintaofei/codeg · 190 tokens

pr-integration-test

Design, implement, and validate Intelligent Terminal integration tests for a target pull request or regression. Use when asked to add PR integration tests, convert a bug fix into E2E coverage, prove existing behavior still works, map tests to the release checklist, or verify E2E reports mark checklist cases complete.

microsoft/intelligent-terminal · 66 tokens