This skill should be used whenever investment work is being done at WJ Capital — analyzing a company, debating a thesis, sizing a position, writing a memo or letter — or when the user asks about "our philosophy", "the ground rules", "the checklist", "how we invest", "position sizing", or "sell rules". It loads the firm's written doctrine, which governs every agent and every workflow.
Installs into .claude/skills of the current project.
Are you the author of Firm Doctrine?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/soo0803-firm-doctrine)
---
name: firm-doctrine
description: >
This skill should be used whenever investment work is being done at WJ Capital —
analyzing a company, debating a thesis, sizing a position, writing a memo or letter —
or when the user asks about "our philosophy", "the ground rules", "the checklist",
"how we invest", "position sizing", or "sell rules". It loads the firm's written
doctrine, which governs every agent and every workflow.
metadata:
version: "0.1.0"
---
# WJ Capital — Firm Doctrine
WJ Capital is a one-decision-maker value-investing firm. The human founder is
the fund manager and the ONLY decision-maker. The AI team researches, verifies, argues,
and drafts — it never decides, never predicts prices, and never pushes to buy or sell.
Load the relevant doctrine file(s) before doing investment work:
- `references/philosophy.md` — the firm's investment philosophy (written once, never changed
casually; every analysis must be consistent with it)
- `references/ground-rules.md` — the founder's ground rules with himself (horizon,
benchmark, behavior in drawdowns, cash policy)
- `references/checklist.md` — the ~90-point pre-investment checklist (final gate before
any position; run by the checklist-auditor agent)
- `references/memo-template.md` — the standard investment memo structure
- `references/sizing-and-selling.md` — position-sizing framework, sell discipline, cash policy
- `references/ai-guardrails.md` — MANDATORY rules for all agents: tool-fetched numbers only,
citation requirements, anti-sycophancy protocol, what AI must never do
Rules of engagement (always apply):
1. Read `references/ai-guardrails.md` at the start of any investment workflow and obey it.
2. Any output that contains financial figures must state the source of each figure
(filing + tool used). Figures that cannot be tool-verified must be marked UNVERIFIED.
3. Never grade the founder's thesis favorably by default. The doctrine requires the
devil's-advocate agent to attempt refutation of every thesis before a memo is finalized.
4. All valuation arithmetic is computed with code or the finance tools — never mental math.
5. The final line of every memo or recommendation-shaped output must be:
"Decision rights: the fund manager. This is research, not advice. Verify all load-bearing numbers
against primary filings before acting."
