Augmented Coding Patterns — Distilled Reference

Short description definitions of the patterns and anti-patterns from Augmented Coding Patterns that zolletta-metaskill's architecture follows or avoids. Skills cite these by name; this file is the local cache so skills don't depend on an external URL at review time.

Source: https://github.com/lexler/augmented-coding-patterns (lexler)

Patterns (followed)

Pattern	Short description	Link
Focused Agent	An agent with a narrow scope follows ground rules more reliably than one juggling many responsibilities.	source
Chain of Small Steps	Break complex goals into small, focused, verifiable steps — each verified before building on it.	source
Feedback Loop	Give AI a clear success signal and permission to iterate autonomously until the goal is reached.	source
Slice for Review	Let the agent deliver large work, then split it into small, coherent, independently reviewable units before human review.	source
Active Partner	Transform the command relationship into two-way dialogue — AI pushes back, flags contradictions, proposes alternatives.	source
Reverse Direction	Present the problem, not your solution — let AI's breadth reveal approaches you haven't considered.	source
Happy to Delete	Treat AI-generated code as disposable exploration — revert freely, start fresh with lessons learned.	source
Shared Canvas	Use markdown files as a shared canvas where humans and AI collaborate on specs, docs, and plans.	source
Constrained Tests	Design tests so that coverage becomes a reliable quality metric — make it impossible to write tests without assertions.	source

Anti-patterns (avoided)

Anti-pattern	Short description	Link
Distracted Agent	One agent doing everything stays shallow and misses what matters.	source
Big Chunk Working	AI generates sweeping changes that are hard to review and easy to merge with hidden damage.	source
Cognitive Overload	Too many open loops, fragmented attention — the team loses a shared picture of what matters.	source
Obsess Over Rules	More rules → more ignored rules; context rot degrades output as the ruleset grows.	source
Flying Blind	Accepting AI-generated code without review — bugs accumulate silently, nobody understands what was built.	source
Unvalidated Leaps	Building on unverified assumptions about the code — each wrong step compounds.	source
Perfect Recall Fallacy	Expecting AI to remember library details from training — give it tools to discover instead.	source
Tell Me a Lie	Forced premises make AI fabricate plausible nonsense to meet an arbitrary requirement.	source
Answer Injection	Putting solutions in questions limits AI to your preconceived approach instead of leveraging its breadth.	source
Silent Misalignment	AI complies instead of flagging contradictions — misalignment grows silently as it builds on wrong interpretations.	source
Sunk Cost	Pushing AI past diminishing returns — code gets messier with each iteration.	source
AI Isolation	Retreating into solo AI work — less pairing, less shared knowledge, weaker collective ownership.	source
AI Slop	Using AI output without adding human judgment — if anyone with your prompt gets the same result, it's slop.	source

How zolletta-metaskill maps to these patterns

Skill / mechanism	Follows	Avoids
Subcommand dispatch (meta-skill → focused subcommand)	Focused Agent	Distracted Agent
One subagent per review area, parallel	Focused Agent, Chain of Small Steps	Distracted Agent, Big Chunk Working
Scripts-first protocol (Phase A → B → C)	Chain of Small Steps, Feedback Loop	Unvalidated Leaps
Separate SUMMARY.md + TODO.md + specialist reports	Slice for Review	Big Chunk Working, Cognitive Overload
cache/_context.md (shared context digest)	—	Cognitive Overload, Obsess Over Rules
"No borderline — emit or suppress, never hedge"	Active Partner	Tell Me a Lie, AI Slop
"Reason to change" test before God class verdict	Active Partner	Unvalidated Leaps, Perfect Recall Fallacy
Coverage cross-check before "missing tests" finding	—	Unvalidated Leaps
Suppress-with-reason convention for Any/mixed	Active Partner	Silent Misalignment
Review-mode (read-only, no fixes)	—	Sunk Cost
Fresh timestamped run folder per review	Happy to Delete	Sunk Cost
Previous review comparison (carry-forward)	Feedback Loop	Flying Blind
Team-facing reports in standard markdown	Shared Canvas	AI Isolation
Deterministic scanner output in reports	—	AI Slop
External review on modified files only	Slice for Review	Flying Blind

Full analysis with evidence and enhancement suggestions: Anti-patterns audit.