Skip to content
logo

Agentic decision workflow

Updated 2026-06-087 min read
#ai
#agent
#github
#adr
#decision-making
#knowledge-base
#automation
#workflow
#human-in-the-loop
Summary

Agentic decision workflow는 에이전트가 GitHub issue, 리서치, ADR, PR, 리뷰 댓글 반영을 자동화하되 사람은 되돌리기 어려운 선택과 trade-off 승인에 집중하는 운영 패턴이다. 실행 기록은 issue/PR에, 결정의 이유는 append-only ADR에, 현재 팀 지식은 mutable wiki에 남기는 3층 구조가 핵심이다.

Key claims

  • Agentic workflow의 목표는 사람을 완전히 제거하는 것이 아니라, 사람이 decision making에 집중하도록 반복 작업과 정리를 agent에게 넘기는 것이다.
  • Atlassian HULA 사례는 좋은 기준점을 준다. agent는 work item을 읽고 plan, code, PR까지 만들지만, engineer가 plan approval, code review, feedback을 담당한다.1
  • GitHub Agentic Workflows는 agent 자동화를 GitHub Actions 안에 넣고 permissions, logs, review boundary를 repository owner가 통제하는 방향을 제안한다. Continuous AI는 CI/CD를 대체하지 않고, 주관적·반복적 repository 작업을 보완한다.2
  • AI coding agent PR 연구는 구조화된 PR 설명이 reviewer 부담을 줄일 수 있음을 보여준다. 하지만 설명만으로 merge가 보장되지는 않고, 코드 품질과 검증이 여전히 중심이다.3
  • ADR은 decision의 context, alternatives, rationale, consequences를 남기는 lightweight 기록이다. Microsoft는 ADR을 append-only log로 다루고, 결정이 바뀌면 기존 record를 수정하지 말고 새 record가 supersede하도록 권장한다.4
  • 따라서 실행 기록은 issue/PR, 결정 이유는 ADR, 최신 팀 지식은 wiki에 둔다. 이 3층을 분리해야 agent가 과거 논쟁을 반복하지 않고 현재 지식과 과거 이유를 함께 읽을 수 있다.

Knowledge capture architecture

PR comments, issue comments, commits, Slack threads, meeting notes, and local agent sessions are all communication artifacts. None of them should be the long-term source of truth by itself. The durable layer should be a central knowledge inbox plus a curated wiki.

LayerPurposeExample trigger
SignalSomething happenedPR merged, issue closed, Slack decision, agent session summary
InboxCandidate knowledge waiting for triageknowledge-inbox/YYYY-MM-DD/*.md
TriageDecide whether the signal is reusablediscard / task log / ADR / wiki update
ADRPreserve why a decision was madeaccepted architecture or policy choice
WikiPreserve current reusable knowledgepattern, concept, playbook, project map

A good default is to run this outside any single product repository. Hermes cron or webhook can watch multiple repositories, collect candidate signals, and open a PR against the central wiki repo. Individual PR comments can still be useful input, but they should not be the primary workflow.

Scope ladder and search/permission rules

Team knowledge should move through a scope ladder rather than through a tool-specific hierarchy.

ScopeCanonical artifactAggregation rule
Personalscratch notes, AI work logspromote only after review
Projectproject context, ADR, requirement, runbooksource of truth stays near code/work
Teamplaybook, reusable workflow, team decision indexsummarize and link to projects
Organizationpolicy, cross-project pattern, customer-level mappublish only reviewed syntheses

Searchability requires structured metadata in addition to prose. Durable pages should include type, scope, project, team, status, visibility, owners, sources, related, updated, and review_after fields. Hybrid retrieval should combine keyword/BM25 for exact identifiers, vector search for paraphrases, metadata filters for scope/status/project, and wikilinks for decision-requirement-code relationships.

Permission is not a presentation concern. Source ACLs from Slack, documents, repositories, or tickets must be carried into metadata at ingest time and enforced before retrieval. A synthesized page cannot be more public than its most restrictive source unless a human explicitly redacts and approves it. Public blog content, internal wiki state, confidential customer material, and restricted personal/HR notes should remain separate collections.

Second Brain and CMDS mapping

Second Brain/PARA is useful because it organizes information around actionability: active projects, ongoing areas, reusable resources, and archives. For organizations, PARA needs additional fields for permissions, owners, source provenance, review cycle, and promotion status.

구요한/Yohan Koo's CMDS vault offers a stronger process language for agentic knowledge operations. The Connect → Merge → Develop → Share lifecycle maps naturally to Hermes workflows:

CMDS stageAgentic knowledge operation
Connectcollect raw signals from Slack, meetings, customer docs, PRs, tickets, and agent sessions
Mergededupe, classify scope, link related project/team/org artifacts
Developpromote candidates into ADRs, requirements, runbooks, wiki pages, or project briefs
Sharepublish reviewed outputs as blog posts, team digests, customer reports, PR updates, or wiki indexes

The cmds-llm-wiki example is especially relevant: raw sources stay in 10. Raw Sources/, the LLM-maintained wiki has index.md, log.md, Wiki/, and Queries/, and Core Context.md acts as a pointer to canonical context files instead of duplicating them. For team and organization knowledge, this pointer-first design reduces drift between project-local facts and central summaries.

Operating loop

단계Agent outputHuman gatePersisted artifact
Repo scanrepo map, constraints, existing conventions목표·제약 확인wiki draft 또는 project brief
Researchrelated docs, examples, risks빠진 맥락 피드백research note, source links
Decision briefoptions, trade-offs, recommendationchoose / defer / rejectproposed ADR
Planningissues, acceptance criteria, task breakdownscope 승인GitHub issues
Implementationbranch, commits, PR, tests큰 방향 변경 승인PR
ReviewClaude/Codex cross-review, CI interpretationmerge decisionPR comments
Knowledge updatewiki synthesis, ADR links, supersession notesaccepted decision 확인wiki page + ADR log

Decision gate rule

Low-risk operations should run automatically. High-impact decisions should stop at a human gate.

Ask a humanLet the agent continue
Public API changeTest addition
DB schema or migrationFormatting, lint, small refactor
Security, auth, privacy, cost impactPR summary and changelog draft
Architecture direction with hard rollbackDocumentation update
ADR status change to acceptedADR draft in proposed state
Product or UX trade-offEvidence collection and comparison

This balances performance better than both extremes. Full manual review wastes time on mechanical work; full autonomy hides irreversible decisions inside generated commits.

Multi-model split

  • Hermes Agent or a similar orchestrator should own state, tools, repository actions, memory, and delivery.
  • GPT-5.5/Codex is a good default implementer and structured reviewer, especially when JSON output or exact file/line findings are needed.5
  • Claude Opus / Claude Code is useful as a design critic and reviewer, especially for architecture, UX, and instruction-following heavy review. Claude Code GitHub Actions can respond to mentions in issues and PRs, while an automation workflow can run without a mention when configured with a prompt.6
  • A second model should not be treated as an oracle. It is a different failure mode. Hermes should still verify diff, tests, CI, and final PR state.

Wiki and ADR division

llm-wiki and ADR solve different memory problems.

LayerTime modelAnswersMutation rule
Issue / PRevent logWhat happened during execution?append comments and commits
ADRappend-only decision logWhy did the team choose this?never rewrite accepted decisions; supersede
Wikimutable semantic layerWhat is true now?update as current understanding changes

A team decision becomes durable knowledge only when the wiki page includes:

  1. decision summary,
  2. context and constraints,
  3. options considered,
  4. decision drivers,
  5. consequences and revisit conditions,
  6. links to ADR, issue, PR, and relevant code paths.

Without context and rejected alternatives, the wiki becomes a conclusion dump. Future agents can repeat the same debate because they cannot see why previous options were rejected.

Connections

  • llm-wiki - mutable semantic knowledge layer that should absorb accepted decision outcomes.
  • github-actions-gcp-cicd - deterministic CI/CD remains the verification layer; agentic workflows should complement it, not replace it.
  • vibe-coding - agentic development is safer when vibe coding is constrained by decision gates, tests, and ADRs.
  • second-brain-rag - decision-aware wiki pages make retrieval more useful because they encode rationale, not just facts.
  • llm-wiki - CMDS-style raw/source separation, index/log discipline, and query filing are useful implementation details for this workflow.

Footnotes

Footnotes

  1. Atlassian, "Human in the Loop Software Development Agents"

  2. GitHub Blog, "Automate repository tasks with GitHub Agentic Workflows"

  3. Watanabe et al., "How AI Coding Agents Communicate"

  4. Microsoft, "Maintain an architecture decision record"

  5. OpenAI Cookbook, "Build Code Review with the Codex SDK"

  6. Claude Code GitHub Actions

Linked from (10)

Comments