More screenshots
Works with the AI Agents you already use
Kill Criteria Thesis Monitor: exit rules before you need them
Write price, fundamental, event and thesis exit rules, then triage news into noise, watch or damaged. Includes an offline checker script.
$9
Kill Criteria Thesis Monitor: exit rules before you need them
Example session with this skill installed
Write exit rules for my AAPL position (average cost 180) and check them against these closes and the news log I keep.
- Read your context and instructions
- Compiled the kill criteria thesis
Run 2026-11-16 on the bundled illustrative fixtures:
as of 2026-11-15 (12 rules)
near P-AAPL-1 AAPL close <= 150.0 close 151.5 (2026-11-12) vs level 150, +1.00% above it
MET P-AAPL-2 AAPL drawdown_high 20 -24.25% from 365-day high, limit -20%, gap -4.25 pts
MET P-MSFT-1 MSFT drawdown_cost 25 -27.50% from cost, limit -25%, gap -2.50 pts
MET P-JPM-1 JPM weight_above 15 weight 17.50% vs 15%, gap -2.50 pts
MET F-NVDA-1 NVDA revenue_growth_below 20 last 2q 13.64, 15.00% vs under 20%, gap -6.36 pts
ok F-ASML-1 ASML gross_margin_below 50 last 1q 54.00% vs under 50%, gap +4.00 pts
MET E-SAP-1 SAP ceo_change may have met 2026-11-12: "the board thanks the outgoin"
ok E-O-1 O dividend_cut no quoted event in the log (not proof it did not happen)
MET T-NVDA NVDA thesis pillar A: 2 watch days in 30d (10-20, 11-10)
MET T-ASML ASML thesis pillar B: 2 watch days in 30d (10-25, 10-26)
ok T-MSFT MSFT thesis no damage, no repeated watch
retired P-OLD AAPL close <= 300.0 skipped
ignored: log row 6 (T-MSFT): damaged without a quote is ignored
7 met of 12. A met rule is a prompt to re-read the thesis and decide, not an order to sell.
Connects securely to your tools. The creator never sees your data.
About this skill
Most sell decisions are made in the worst hour: the stock is down, the headline is loud, and any reason sounds good. This skill moves the decision earlier. It teaches your agent to write exit rules while calm, in terms an observation can settle, and to review news against them without selling a sound business on one bad day.
What it does
Your agent drafts kill criteria per position across four families (price, fundamental, event, thesis with one to three pillars), shows them for your approval, then sorts incoming filings and headlines per pillar into noise, watch or damaged. A thesis counts as broken on one damaged fact, or when the same worry shows up on two different days within 30 days. A met rule is a prompt to re-read the thesis and decide, never an automatic sell.
What you get
| Path | Holds |
|---|---|
| SKILL.md | rules, criteria table, verdict table, write-monitor-review method |
| templates/thesis-card.md | one-screen thesis card per position |
| templates/kill-criteria.md | criteria sheet the script reads |
| templates/review-log.md | verdict log the script reads |
| references/writing-good-criteria.md | sizing levels, quarters and pillars; weak vs good rewrites |
| references/news-triage.md | the sieve, the sort, 8-K items, failure modes |
| references/decision-journal.md | recording decisions, 30 and 90 day marks |
| scripts/check_criteria.py | offline checker: price, fundamental and two-watches rule |
Who it's for
Self-directed investors with a handful to a few dozen positions, and builders of portfolio assistants that should propose exit rules and read news against them.
Scope
Covered: price, drawdown, weight, quarterly fundamental, CEO, guidance and dividend rules, thesis pillars, news triage, a decision journal. The script runs offline on files you supply (Python 3.9+, standard library).
Not covered: fetching prices, filings or news; the script counts verdicts you log and refuses ones with no quote. Crypto and funds have no filings, so only price and sizing rules apply.
Not investment advice.
The script does not read news or filings or fetch prices; it counts verdicts you log and flags closes older than 7 days. Fundamental rules need the quarterly figures supplied. Event rules are closed questions about a quoted passage, not keyword matches. Not investment advice.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 day ago
Requires
Permissions
Offline. Reads only the files you pass to the script; no network, no writes.