A directional intelligence skill for AI agents.
Skate where the puck is going, but show your evidence.
Most people ask: What is happening?
/sktpg asks: What is this becoming?
Better: What does this force next?
Best: What becomes obvious 6–18 months from now that is still non-obvious today?
/sktpg forces you to stop thinking about "what is happening now?" and start thinking in:
- Weak signals — early motion before consensus forms
- Hype vs real motion — narrative separation with evidence
- Direction of travel — vector, not point
- Incentive maps — who is pulled, who is squeezed, who is forced to move
- Bottleneck maps — current bottleneck → next bottleneck → who profits
- Second-order questions — 10× cheaper? regulated? default infra?
- 6–18 month forecasts — base / bull / bear / wild card, all falsifiable
- Action maps — what to do in 24h / 7d / 30d / before consensus
- Tracking signals — green / yellow / red scoreboard
- The SKTPG Thesis — compact 6-line verdict
Every conclusion is tagged by evidence strength: Confirmed · Likely · Speculative · Contradicted · Unknown
SKTPG works on any platform that supports AI agent skills — Hermes, Claude Code, Cursor, Copilot, Codex, etc.
hermes skill install https://github.com/New1Direction/SKTPGClone the repo, then link the skill:
git clone https://github.com/New1Direction/SKTPG ~/.claude/skills/sktpgOr use the plugin manifest if you publish via marketplace.
Clone the repo and symlink SKILL.md into your agent's skills directory:
git clone https://github.com/New1Direction/SKTPG ~/sktpg
# then symlink SKILL.md or the whole folder into your agent's skills pathThen use:
/sktpg What's the real signal behind the AI agent framework explosion?
/sktpg https://github.com/some-project
/sktpg API inference prices just dropped 80%
The skill ships with verification cases in references/:
| Test | Input | Expected score |
|---|---|---|
| Hype-only | "This AI agent will replace all workers." | 0–20 (noise) |
| Real weak signal | "Small teams replacing QA with browser agents, humans for final approval" | 41–70 (watchlist) |
| Bottleneck shift | "API provider cuts inference prices by 80%" | 60–85 (actionable) |
| Repo weak signal | Repo with prompts/mocks, no runtime proof | Identify scaffold vs tool |
| Protocol incentive | L2 changes fee distribution favoring stakers | Map forced moves |
| Category | Points | Question |
|---|---|---|
| Signal strength | 20 | Real weak signals beyond hype? |
| Incentive alignment | 20 | Are actors forced to move? |
| Bottleneck shift | 15 | Real constraint being removed? |
| Second-order upside | 15 | Unlocks new markets/workflows? |
| Timing advantage | 15 | Action now matter before consensus? |
| Falsifiability | 10 | Thesis trackable and killable? |
| Personal edge | 5 | User has plausible advantage? |
0–20 Noise · 21–40 Interesting · 41–60 Watchlist · 61–80 Actionable · 81–100 Urgent opportunity
- Do not overpredict — directional theses with falsification criteria
- Do not summarize only — explain what it forces next
- Do not confuse attention with motion — usage > hype, adoption > funding, workflow > demo
- Do not ignore bottlenecks — current bottleneck AND next bottleneck
- Do not skip action — always answer "what should I do before this becomes obvious?"
MIT