Context Vigilance
Search/

The Hook

Treat context with the same vigilance as code — and the context becomes the parent to the code.

Most coding agents fail not because the code is hard, but because the context is missing. You hand the agent a problem; it doesn't know your conventions, your prior decisions, the mistakes you made last quarter. So it guesses.

Treat context with vigilance — versioned, reviewed, cross-linked — and something flips. Code becomes regenerable. A bug fix compresses from days to minutes. Migrating across languages goes from "we'll plan a quarter for that" to "we did it Tuesday afternoon." The agent didn't get smarter. It finally has the ground it needed to stand on.

Try:

Browse the corpus What is this practice?

The Practice

Eight folders. Three cognitive modes.

Every file is written for three readers: the human editing it, the agent loading it as context tomorrow, and the public reader who lands on it cold. Frontmatter is a contract. Cross-links are how attention navigates a tree too big for any one head.

Modes
  • Prep Deciding what to build before we build it.
  • Reflection Codifying what works and what doesn't after we built it.
  • Journey Capturing research and the painful debugging paths along the way.

Experimental tier

Four proposed folders the practice is actively trialing. Their shape is deliberately not yet enforced — divergence between repos is expected and surfaced, not normalized. This is the on-ramp every canonical folder took: plans/ and agent-skills/ both started here.

  • loops/ Recurring operational loops — scope, cadence, per-iteration procedure, exit conditions.
  • handoffs/ End-of-session state capture so the next session — human or agent — resumes without re-derivation.
  • decisions/ Isolated artifacts of decisions made — what, when, why, and the alternatives passed over. Kin to ADRs.
  • contracts/ Not suggestions: constitution rules agents must always follow, or ironclad API-level data-flow contracts.

In practice, also

Real-world context-v has a long tail beyond the canonical eight. Counted across the corpus, the most common extra surfaces:

  • profiles/ 54
  • narratives/ 53
  • sitemap/ 45
  • models/ 9
  • habits/ 7
  • inquiry/ 3
  • workflow/ 2
  • loops/ 1

In isolation each makes sense; some are domain-specific and correct where they are. The discipline tolerates the long tail — that's part of the proof — and promotes what earns it: assimilation, not enforcement.

The Proof

1001 living docs. 40 projects.

One small agency's actual work over 18 months — not a theory paper. 107 mature. 540 started. 354 stubs we'll fill out with agent help. Some are specs you can fork. Some are reminders born after the third time we made the same mistake. Some are explorations that ended in "we decided not to pursue" — and those are valuable too.

Browse all 1001 docs → Search the corpus

The Invitation

This isn't our IP.

The directory pattern, the file format, the practice — they're an open specification, in the spirit of the original Hyperloop paper. We thought of this; here's how it works; someone please build it, fork it, improve on it, or adopt it as-is. Schema and tooling are MIT.

For developers

Fork the kit.

The collator script, the schema, the splash you're reading — all on GitHub. MIT-licensed. Treat it as a starter, not a framework.

GitHub →

For practitioners

Browse the corpus.

1001 files of working notes from 40 projects. Search, sort by repo, lift what's useful. The vocabulary is yours to use.

Open the corpus →

For everyone else

Read the spec.

The practice, the why, and where it sits among agents.md, llms.txt, and spec-kit. Fellow travelers, not competitors.

The spec →

— Build with us, build alongside us, or build your own.