# Sf Edge Watch

> Diagnose whether a strategy's edge is alive, softening, decaying, or dead — by comparing forward-test vs backtest profit factor / win rate and 7d vs 30d vs 90d recency. Use when the user asks "is this strategy still working / is the edge alive / is X dying / health-check this strategy". Read-only.

- **Type:** Skill
- **Install:** `agentstack add skill-daviddme-claude-skills-sf-edge-watch`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [daviddme](https://agentstack.voostack.com/s/daviddme)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [daviddme](https://github.com/daviddme)
- **Source:** https://github.com/daviddme/claude-skills/tree/main/SF-Skills/sf-edge-watch

## Install

```sh
agentstack add skill-daviddme-claude-skills-sf-edge-watch
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# sf-edge-watch

Edges decay. This skill tells you whether one is still alive, by checking
two independent signals:

**1. Forward-vs-backtest divergence.** If forward-test profit factor has
dropped below half of backtest PF, the live edge has cratered relative to
the curve fit. Same logic on win rate.

**2. Recency collapse.** Annualize the 7-day, 30-day and 90-day returns
and compare. If the annualized 7d has crashed to less than half of
annualized 30d, the regime has turned hostile.

Each fired flag is listed. Status is then assigned:

| Status     | Meaning                                                   |
| ---------- | --------------------------------------------------------- |
| `alive`    | No flags. Keep running.                                   |
| `soft`     | One mild flag (7d small drawdown). Watch it.              |
| `decaying` | 2 flags, or one moderate flag. Consider trimming.         |
| `dead`     | 3+ flags, or forward PF under half of backtest. Pull it.  |

## When to use

- "Health-check this strategy"
- "Is `trendhoo-v5-by-daviddtech-2543` still working?"
- "Is the edge dying on my deployed bots?"
- Called as a subprocess by `sf-prune` and `sf-loop-pruner`

## How to run

```
# Single strategy
python sf-edge-watch/watch.py STRATEGY_ID_OR_SLUG [--json]

# Audit ALL strategies tied to your deployed bots
python sf-edge-watch/watch.py --deployed [--json]
```

## Output

For one strategy: status, flags list, and the underlying metric comparison
(forward PF vs backtest PF, etc.).

For `--deployed`: a table — bot name, strategy, status, flag count — sorted
worst-first.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [daviddme](https://github.com/daviddme)
- **Source:** [daviddme/claude-skills](https://github.com/daviddme/claude-skills)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** yes
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-daviddme-claude-skills-sf-edge-watch
- Seller: https://agentstack.voostack.com/s/daviddme
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.