self-test

self-test is a command for coding agents from technicalpickles/pickled-claude-plugins. It costs 12 tokens per session (761 once invoked), scanned A, original, MIT.

Run behavioral tests to validate stay-on-target prompts.

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/technicalpickles/pickled-claude-plugins/self-test
Clone the repo
git clone --depth 1 https://github.com/technicalpickles/pickled-claude-plugins

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for self-test

README.md
[![agentmods](https://agentmods.dev/badge/commands/technicalpickles/pickled-claude-plugins/self-test.svg)](https://agentmods.dev/commands/technicalpickles/pickled-claude-plugins/self-test)
Your own site
<a href="https://agentmods.dev/commands/technicalpickles/pickled-claude-plugins/self-test"><img src="https://agentmods.dev/badge/commands/technicalpickles/pickled-claude-plugins/self-test.svg" alt="Measured on agentmods" height="20"></a>
Per session 12 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 761 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin unknown 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 $0.00012 $0.00761
Opus 5 $0.00006 $0.00380
Sonnet 5 $0.00002 $0.00152
Haiku 4.5 $0.00001 $0.00076

Measured today against content hash 867672b1b8f4, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

self-test 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 today.

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.

plugins/stay-on-target/commands/self-test.md · 122 lines

How it starts

The opening of the file, as written. The whole thing — 122 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Self-Test

Run behavioral tests comparing baseline Claude behavior against stay-on-target enhanced behavior.

Process

  1. Load scenarios from test/scenarios/*.md
  2. For each scenario: a. Parse frontmatter for git_ref, cwd, criteria b. Resolve ${PLUGIN_ROOT} in cwd to actual path c. Compose the stay-on-target prompt (read all prompt modules) d. Dispatch TEST subagent (baseline) - just the user message, no prompt injection e. Dispatch TEST subagent (treatment) - with stay-on-target prompt prepended f. Load rubric and reviewer prompt g. Dispatch REVIEWER subagent to grade baseline response h. Dispatch REVIEWER subagent to grade treatment response i. Record scores
  3. Write results to test/results/YYYY-MM-DD-HHMMSS.md
  4. Print summary showing baseline vs treatment delta

Composing the Prompt

Read and concatenate:

  1. prompts/_base.md
  2. prompts/behaviors/*.md (in sorted order)

This mirrors what hooks-handlers/session-start.sh does.

Context Isolation

Critical: Test subagents must NOT be aware they are being tested. This prevents meta-awareness where the subagent recognizes it's in a test scenario and behaves differently.

Isolation techniques:

  1. Use role-play framing ("You are a coding assistant...")
  2. Provide complete simulated environment (don't reference real context)
  3. Set working directory to test fixture (tmp/bktide), not real workspace
  4. Don't mention "test", "scenario", "baseline", "treatment" in prompts

TEST Subagent Prompt (Baseline)

ROLE: You are a coding assistant helping a developer.

ENVIRONMENT:
- Working directory: {cwd}
- This is the start of a new conversation

USER REQUEST:
{user_message}

YOUR TASK:
Respond to the user's request. This is a real development session - help them effectively.

TEST Subagent Prompt (Treatment)

ROLE: You are a coding assistant helping a developer.

ENVIRONMENT:
- Working directory: {cwd}
- This is the start of a new conversation

ADDITIONAL INSTRUCTIONS:
{composed_stay_on_target_prompt}

---

USER REQUEST:
{user_message}

YOUR TASK:
Respond to the user's request following the additional instructions above. This is a real development session - help them effectively.

Read the full file on GitHub · 122 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. today First seen · 122 lines · 12 tokens per session scan A 867672b1b8f4

Subscribe to this mod's changes

self-test is a command published in the GitHub repository technicalpickles/pickled-claude-plugins (10 stars, last pushed today), licensed MIT. It adds 12 tokens to every session and 761 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 2026-09-03.