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.
npx skills add OutlineDriven/outline-driven-development --skill corroborate-by-independent-rerunsgit clone --depth 1 https://github.com/OutlineDriven/outline-driven-developmentWrote 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.
[](https://agentmods.dev/skills/outlinedriven/outline-driven-development/corroborate-by-independent-reruns)<a href="https://agentmods.dev/skills/outlinedriven/outline-driven-development/corroborate-by-independent-reruns"><img src="https://agentmods.dev/badge/skills/outlinedriven/outline-driven-development/corroborate-by-independent-reruns/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/outlinedriven/outline-driven-development/corroborate-by-independent-reruns"><img src="https://agentmods.dev/badge/skills/outlinedriven/outline-driven-development/corroborate-by-independent-reruns.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00036 | $0.00767 |
| Opus 5 | $0.00018 | $0.00383 |
| Sonnet 5 | $0.00007 | $0.00153 |
| Haiku 4.5 | $0.00004 | $0.00077 |
Grade A, and why
corroborate-by-independent-reruns 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 6d ago.
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.
How it starts
The opening of the file, as written. The whole thing — 42 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Corroborate by independent reruns
Contract
| Field | Bound contract |
|---|---|
| Trigger | A candidate patch or answer needs independent corroboration before it is trusted |
| Authority | Reversible local: writes only the named corroboration report; rollback is discarding it if rejected. No remote mutation. Each run mutates only its own isolated sandbox, discarded after capture. |
| Side effect | Runs N isolated sandboxed attempts and returns the majority patch plus its agreement count; byte identity is a corroboration signal, not a correctness oracle; no-change runs never carry consensus |
| Done | Either a strict majority of launched runs produced byte-identical output and it is returned with its count, or no consensus is reported |
Inputs
- The candidate patch or answer to corroborate.
- N, the number of isolated runs to launch (N >= 1).
- The sandbox execution command or task prompt each run performs independently from a clean baseline.
- Optional: a baseline output to classify no-change runs against.
Procedure
- Receive the candidate, N, and the sandbox execution command. Bound scope to N isolated runs; do not widen to additional runs or mutate the shared working tree. Done when: scope is bounded to exactly N isolated runs with no shared-tree mutation.
- Launch N isolated sandboxed attempts, each starting from the same clean baseline and running the same task without access to any other run's output. Done when: N isolated runs are launched, each from a clean baseline with no cross-run access.
- Capture each run's full output bytes. Done when: every launched run's full output bytes are captured.
- Classify each output: a no-change run whose output is byte-identical to the baseline or empty is excluded from consensus and never carries consensus. Done when: every output is classified as candidate or no-change, with no-change runs excluded from consensus.
- Group the remaining outputs by byte identity. If a single byte-identical group is a strict majority of the N launched runs (count > N/2), return that output with its agreement count. Done when: outputs are grouped by byte identity and either a strict-majority group is identified or no majority exists.
- If no byte-identical group reaches a strict majority, report no consensus. Done when: no-consensus is reported with the per-run output groups, or the strict-majority output is returned with its count.
- Treat byte identity as a corroboration signal only, never as a correctness oracle; do not assert the majority output is correct. Done when: the result is presented as a corroboration signal with no correctness claim.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 6d ago Changed · -26 tokens per session 0b9f8f17cb7c
- 8d ago First seen · 42 lines · 62 tokens per session scan A de424bcff935
corroborate-by-independent-reruns is a skill published in the GitHub repository OutlineDriven/outline-driven-development (52 stars, last pushed 6d ago), licensed Apache-2.0. It adds 36 tokens to every session and 767 once invoked, about $0.0002 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.
Other skills, from other repositories
libfuzzer
Use when asked to build, run, or triage a coverage-guided C/C++ fuzz campaign on the libFuzzer or AFL++ engine. Not for harness design: use fuzz-harness-writing.
strict-validation-setup
Use when a user invokes a strict-mode validation or verifiable-goals loop setup. Bootstraps strict-mode tooling and per-task GOALS.md scaffolding so an agentic loop can self-verify. Don't use for remote, credential, publish, deploy, or irreversible changes.
validation-first-driven
Use when building protocols, workflows, concurrent systems, or lifecycle-heavy state that needs explicit states, transitions, and temporal properties. Defines the state machine, encodes invariants in types, and for high-risk designs runs a TLA+ or Alloy model checker. Not for encoding domain models in types — use…
fuzzing
Use when planning an end-to-end fuzzing program for a project: engine and target selection, corpus management, and CI or nightly wiring. Not for harness design: use fuzz-harness-writing. Not for engine operation or triage: use libfuzzer. Not for sanitizer flag reference: use sanitizers.
oss-fuzz
Use when enrolling a project in OSS-Fuzz, running its helper workflow locally, or reproducing an OSS-Fuzz report. Not for remote, credential, publish, deploy, or irreversible changes.
testing-handbook-generator
Use when the user asks to discover, generate, refresh, or validate skills from the Trail of Bits Testing Handbook or appsec.guide. Not for tasks that require source or remote-system changes.