Cross-Source Theme Analysis#

16 sources, 8 tools, 2 standards, 3 methodologies, 1 practitioner account, 2 skill/eval resources. Here are the themes that appear across 3+ sources independently — not because they reference each other, but because they converged on the same ideas.

Note: This analysis was originally written against 11 sources. The 5 newest sources (Paperclip, Spec Kit, BMad Method, Anthropic Eval Guide, Promptfoo) strengthen existing themes — particularly Theme 3 (human-in-the-loop spectrum) and Theme 7 (evaluation). A full refresh is recommended when the wiki reaches 20+ sources.

Theme 1: Context Is King (9/11 sources)#

The single most repeated idea across the wiki. Almost every source says some version of “getting the right context loaded matters more than anything else.”

The consensus: Don’t load everything. Load the right thing at the right time. Every tool that scales has independently arrived at some form of progressive disclosure.

Theme 2: Composition Over Monoliths (8/11 sources)#

No one builds monolithic agents or skills. Everyone decomposes.

The consensus: Small, focused, composable. The YouTube pipeline (6 stages) vs. a single “make a video” skill is the clearest illustration. This is the strongest convergence in the wiki.

Theme 3: The Human Stays in the Loop — But How Much? (7/11 sources)#

Every source addresses the human-agent boundary, but they disagree on where to draw it.

The spectrum: From Scion (always interactive) to Kiro (hours of autonomy) to PAI (self-modifying). No consensus. Claude Code’s approach — making it a configurable dial — may be the most pragmatic.

Theme 4: Skills Are Evolving Into a Standard (6/11 sources)#

The concept of “a reusable unit of agent capability” is converging on a common format.

The evolution: Fabric Patterns → Agent Skills Standard → Claude Code Skills → Pipelines + Evaluation. The trajectory is clear: from simple prompt files to a full lifecycle (design → build → test → deploy → evaluate).

Theme 5: Memory Is the Next Frontier (6/11 sources)#

How agents persist and accumulate knowledge is the most actively divergent area.

The tension: Persistent memory compounds value (PAI, Kiro, wiki pattern) but also compounds errors (stale learnings, wrong memories). Scion’s fresh-start approach avoids this but wastes re-discovery. No one has solved the “memory hygiene” problem — how to keep accumulated knowledge accurate over time. The llm-wiki-pattern’s “lint” operation (health-check for contradictions) is the closest thing to a solution.

Theme 6: Git as Universal Substrate (6/11 sources)#

Git isn’t just version control — it’s the coordination layer, isolation mechanism, and review process.

The consensus: Git is the one coordination mechanism everyone agrees on. No one is building a custom protocol. But git only works for text-shaped artifacts — the open question is what coordinates non-code knowledge.

Theme 7: Evaluation Is the Weakest Link (4/11 sources)#

The gap between “building agents” and “knowing if they work” is acknowledged but unsolved.

The gap: Everyone knows evaluation matters. Almost no one does it rigorously. The Caparas article provides the methodology; the Skills Pipeline shows inline testing; but no tool has an integrated eval pipeline. This is the biggest opportunity in the ecosystem.

Theme 8: Open Standards Are Winning (5/11 sources)#

Proprietary approaches are losing to open protocols.

The trend: MCP (tools/data) and Agent Skills (capabilities) are emerging as the two-layer open substrate. Fabric’s patterns are the community-driven content layer. Proprietary approaches (Kiro Powers) exist but are converging toward open standards.

Theme Matrix#

How many sources support each theme:

See Also#

• key-insights-agentic-landscape
• ten-pillars-evidence-map
• multi-agent-orchestration
• context-management