How do I install the Challenge skill?

Run npx skillmds@latest add agentii-ai/challenge in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Challenge safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Challenge?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Challenge free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Challenge?

agentii-ai (@agentii-ai) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Challenge by agentii-ai — Agent Skill

Name: Challenge
Author: agentii-ai

Challenge

Adversarial verification of research theses — cross-run/cross-thesis contradiction via the entity index, pre-mortem (Klarman/Kahneman: assume the loss already happened, reverse the path), inversion (Munger: invert, always invert), and wrong_if falsifiability review. ≈50-finding cap sorted by severity; content-derived finding IDs; incremental scoping with three backstops; lifecycle hooks at pillar-complete / thesis-complete / pre-reduction.

agentii-ai 3952034 2 files · 2.7 KB Updated Sep 17, 2026

File contents

agentii.challenge

The buyer-side IC verb: challenge the thesis. Not audit — that name was taken three times (audit-xls, quality-audit.yaml, G3) — challenge is the thing that happens in an investment committee.

Method bodies (the four angles)

Cross-run contradiction — the entity index (reduce_journals.build_entity_index): same (entity, metric, period) with different values. Same retrieved_at → true contradiction; different → suspected restatement. The validator never auto-resolves — candidates go to the IC agenda.
Pre-mortem — assume the loss already happened; reconstruct the path backwards. Which pillar failed first? What data would have to be wrong?
Inversion — what would make this thesis a sell? What would the bear's best case look like? What would falsify the conviction, not just the claim?
wrong_if falsifiability — is each falsifier mechanically checkable (metric + threshold + source)? A prose falsifier is not a falsifier (Q8-4).

Economics (Q40/Q68)

Findings: ≈50 cap, severity-sorted; overflow aggregates by category.
Finding IDs are content-derived (hash(entity+metric+period+gap_type)) — unchanged re-runs yield byte-identical IDs (eval corpus comparability).
Scope is incremental (artifacts whose pins/as_of changed since last run) with three backstops: every Nth converge / constitution MINOR-MAJOR / subscription change (new subscriptions create new comparison pairs — Q9).
Lifecycle hooks: pillar-complete (cheap, early) / thesis-complete (final line) / pre-reduction (claims entering the portfolio are adversarially verified). Unchallenged claims never enter the knowledge base (Q64).

Invocation

python3 scripts/challenge.py --nth-converge 10

---name: challengedescription: Adversarial verification of research theses — cross-run/cross-thesis contradiction via the entity index, pre-mortem (Klarman/Kahneman: assume the loss already happened, reverse the path), inversion (Munger: invert, always invert), and wrong_if falsifiability review. ≈50-finding cap sorted by severity; content-derived finding IDs; incremental scoping with three backstops; lifecycle hooks at pillar-complete / thesis-complete / pre-reduction.---# agentii.challengeThe buyer-side IC verb: *challenge the thesis*. Not `audit` — that name was takenthree times (audit-xls, quality-audit.yaml, G3) — challenge is the thing thathappens in an investment committee.## Method bodies (the four angles)1. **Cross-run contradiction** — the entity index (`reduce_journals.build_entity_index`): same `(entity, metric, period)` with different values. Same `retrieved_at` → true contradiction; different → suspected restatement. **The validator never auto-resolves** — candidates go to the IC agenda.2. **Pre-mortem** — assume the loss already happened; reconstruct the path backwards. Which pillar failed first? What data would have to be wrong?3. **Inversion** — what would make this thesis a sell? What would the bear's best case look like? What would falsify the conviction, not just the claim?4. **`wrong_if` falsifiability** — is each falsifier mechanically checkable (metric + threshold + source)? A prose falsifier is not a falsifier (Q8-4).## Economics (Q40/Q68)- Findings: ≈50 cap, severity-sorted; overflow aggregates by category.- Finding IDs are content-derived (`hash(entity+metric+period+gap_type)`) — unchanged re-runs yield byte-identical IDs (eval corpus comparability).- Scope is **incremental** (artifacts whose pins/`as_of` changed since last run) with three backstops: every Nth converge / constitution MINOR-MAJOR / subscription change (new subscriptions create new comparison pairs — Q9).- Lifecycle hooks: pillar-complete (cheap, early) / thesis-complete (final line) / pre-reduction (claims entering the portfolio are adversarially verified). **Unchallenged claims never enter the knowledge base (Q64).**## Invocation```bashpython3 scripts/challenge.py --nth-converge 10```

Frequently asked questions