> ## Documentation Index
> Fetch the complete documentation index at: https://docs.brik64.tech/llms.txt
> Use this file to discover all available pages before exploring further.

# Anti-Patterns

> Patterns agents should avoid in BRIK64 docs and PCD.

BRIK64 agent work fails when the artifact becomes harder to inspect.

Avoid these patterns.

## Invented Vocabulary

Do not invent monomer names, file standards, command flags or evidence labels. Use the documented catalog and CLI reference.

## Hidden Behavior

Do not hide important behavior in helper functions without explaining the circuit role. If the helper is necessary, explain its inputs, output and failure behavior.

## Open Branches

Do not omit blocked cases or return different result kinds from different branches.

## Unsupported Syntax

Do not write typed-argument syntax, arrow returns or symbolic EVA expressions as executable PCD unless the release documentation supports that syntax.

## Unauthorized Agent Writes

Do not modify `AGENTS.md` through `brik64 init`. Use the documented skill
install path or host agent environment and require explicit review before
writing. See [Use The BRIK64 Skill](https://docs.brik64.tech/agents/install-skill.md).

## Overstated Results

Do not use local checks as broad application claims. Pair results with command output, files changed and evidence status.

## Internal Vocabulary In Product UI

Do not use internal engine jargon in ordinary user-facing product copy unless
the page is explicitly an expert developer console. Prefer product-domain
language:

| Prefer | Instead of exposing raw internals |
| - | - |
| bounded logic, domain rule, validation contract, review artifact | raw PCD identifiers in primary UI labels |
| local trace, trace history, audit trail, review state | internal ledger jargon in primary UI |
| checked for this scope, traceable | language that implies external certification |
| composed rule set, system rule set | raw polymer jargon in primary UI |

Raw PCD names, hashes, `.brik` paths and command output belong in expandable
developer diagnostics, logs, reports or agent-readable evidence, not in primary
application labels.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.