{
  "markdown": "# Software Engineering Handbook\n\nA public, risk-scaled software-engineering handbook and sparse\nmixture-of-experts skill router for humans, teams, and AI coding agents.\n\n**11 supported harness surfaces · 5 chapters · 8 reviewed providers · 74 normalized purposes · MIT**\n\n> **Context-first:** agents discover a tiny descriptor at startup and load only\n> the handbook and expert sections required by the current task.\n\n## Table of contents\n\n- [Quick start](#quick-start)\n- [Usage](#usage)\n- [Features](#features)\n  - [Engineering coverage](#engineering-coverage)\n  - [Complete capability inventory](#complete-capability-inventory)\n  - [Routing and package utilities](#routing-and-package-utilities)\n- [How it works without filling the context window](#how-it-works-without-filling-the-context-window)\n- [Directory roles](#directory-roles)\n- [Install on supported agents](#install-on-supported-agents)\n  - [Claude App](#claude-app)\n  - [Claude Code](#claude-code)\n  - [Antigravity](#antigravity)\n  - [Codex App](#codex-app)\n  - [Codex CLI](#codex-cli)\n  - [Gemini CLI](#gemini-cli)\n  - [Kimi Code CLI](#kimi-code-cli)\n  - [OpenCode](#opencode)\n  - [Hermes Agent](#hermes-agent)\n  - [Pi](#pi)\n  - [Oh My Pi](#oh-my-pi)\n- [Future harness support](#future-harness-support)\n- [Updating](#updating)\n- [Repository contents](#repository-contents)\n- [Validate and contribute](#validate-and-contribute)\n- [Safety and repository boundary](#safety-and-repository-boundary)\n- [Upstream attribution](#upstream-attribution)\n- [License](#license)\n\n## Quick start\n\nChoose the scoped installation procedure for your host under\n[install on supported agents](#install-on-supported-agents). Each procedure\ninstalls or retains the complete package: portable skill, handbook, registry,\nand the applicable host adapter. The canonical skill references sibling\nhandbook sources, so a skill-only copy is incomplete.\n\nReview the repository and pin a trusted tag or commit when the host supports\nimmutable references. Start a new agent session after installation.\n\n## Usage\n\nYou normally do not need a special command after installation. Ask for the work\nitself:\n\n```text\nFix this concurrency bug and prove the reproduction no longer fails.\nReview this pull request for correctness and security risks.\nPlan the smallest safe implementation of this API change.\nSimplify this module without changing its observable behavior.\nInvestigate this performance regression with a fixed benchmark protocol.\n```\n\nThe router classifies risk, reads repository-specific rules, chooses one primary\nchapter, and activates only the smallest compatible expert set. Project and user\ninstructions remain authoritative over generic handbook defaults.\n\nFor explicit use, ask the agent to “apply the software-engineering handbook” or\ninvoke `software-engineering-handbook` through the harness's skill UI or slash\ncommand.\n\n> **One installation is enough.** Reviewed provider skills are optional\n> compatibility inputs, not dependencies. The handbook includes complete\n> internal fallbacks and does not require Superpowers, Ponytail, Agent Skills,\n> Spec Kit, OpenSpec, BMAD, or another workflow pack to be installed.\n\n## Features\n\n- Five portable chapters covering lifecycle governance, architecture and code\n  quality, testing and debugging, Git/CI/CD/security, and atomic delivery.\n- Risk tiers from localized reversible work through critical irreversible work.\n- A machine-readable registry containing 8 reviewed providers, 11 provider\n  groups, and 90 provider-qualified originals plus 4 handbook-native\n  capabilities, normalized into 74 purposes.\n- Deduplicated MoE routing: each original belongs to exactly one purpose.\n  Equivalent skills are ordered alternatives, so only one original or fallback\n  can load for that purpose.\n- Bounded expert adaptations from Superpowers, Ponytail, Addy Osmani's Agent\n  Skills, Understand Anything, autoresearch, GitHub Spec Kit, OpenSpec, and\n  BMAD Method.\n- Trusted-original preference: use a compatible installed original skill when\n  available; otherwise load the bounded internal fallback, never both.\n- Built-in context and usage economics: degradation diagnosis, safe compression\n  and handoff, recurring harness-surface audits, progressive disclosure,\n  cache-aware reuse, bounded tool/agent fan-out, risk-scaled model choice, and\n  honest total-task token and cost evidence.\n- Governed procedure and skill improvement: causal triggers, bounded candidates,\n  representative development and held-out cases, hard safety gates, independent\n  approval for high-risk changes, rejected-candidate evidence, and rollback.\n- Built-in Ponytail engineering behavior: comprehension-first minimal\n  implementation, root-cause placement, over-engineering review and audit, and\n  shortcut-debt reporting. Ponytail's intensity modes, help card, benchmark\n  scoreboard, hooks, and persistent mode state remain intentionally unbundled.\n- Explicit authority, evidence, conflict, failure, rollback, and context-budget\n  rules.\n\nThis is guidance and routing—not an autonomous deployment system. It does not\nsilently install dependencies, run hooks, publish, release, deploy, access\nsecrets, or grant itself authority.\n\n### Engineering coverage\n\nThe handbook covers the software lifecycle, not just code generation. These\nchapter summaries describe guidance and controls, not automatically executed\nservices:\n\n| Chapter | Features and responsibilities |\n|---|---|\n| [Lifecycle and governance](handbook/software-engineering/01-lifecycle-governance.md) | Scope, stakeholders, requirements, ownership, decision authority, traceability, change control, operation, deprecation, and retirement |\n| [Architecture and code quality](handbook/software-engineering/02-architecture-code-quality.md) | Boundaries, contracts, data and error models, dependency choices, reliability, resource behavior, performance, maintainability, and implementation discipline |\n| [TDD, testing, and debugging](handbook/software-engineering/03-tdd-testing-debugging.md) | Test strategy, red–green–refactor, discriminating reproductions, root-cause investigation, deterministic verification, regression prevention, and completion evidence |\n| [Git, CI/CD, and security](handbook/software-engineering/04-git-ci-cd-security.md) | Coherent commits, protected integration, exact-revision review, CI gates, secrets, supply-chain controls, artifact provenance, authorized release, recovery, and vulnerability response |\n| [Agile atomic delivery](handbook/software-engineering/05-agile-atomic-delivery.md) | Usable vertical slices, sequencing, bounded work, coordination, feedback, Definition of Done, and evidence-bearing handoff without mandating a sprint length or tracker |\n\n### Complete capability inventory\n\nThe inventory below covers **all 90 registered provider-qualified originals**\nand **all 4 handbook-native capabilities**. They resolve to **74 normalized\npurposes: 9 shared-purpose groups, 61 distinct originals, and 4 native purposes**.\nCounts describe routing identities, not 94 independent workflows to run together.\n\nThe [capability registry](skills/software-engineering-handbook/experts/registry.json)\nowns descriptors and provider mappings; the\n[purpose catalog](skills/software-engineering-handbook/experts/purposes.json)\nowns normalization and alternative ordering. This README is a browsing index,\nnot a second routing policy. Capability identifiers are lookup names, not\npromises that a host exposes matching slash commands.\n\n#### Handbook-native capabilities\n\nAll four are included in the\n[context-economics module](skills/software-engineering-handbook/experts/context-economics.md):\n\n| Purpose | Capability |\n|---|---|\n| `context-degradation-diagnosis` | Diagnose lost, poisoned, distracting, confused, or conflicting context before choosing a mitigation |\n| `context-compression-continuity` | Compact or hand off long-running work while preserving exact task state and recoverability |\n| `harness-context-audit` | Audit recurring instruction, skill, tool, memory, and history overhead before authorized cleanup |\n| `context-usage-economics` | Reduce context, token, model, tool, and agent cost without weakening task success or evidence |\n\n#### Shared purposes\n\nThese nine groups normalize overlapping originals. The router uses the first\nmatching trusted original in the catalog's declared order, or the one complete\ninternal fallback; it does not combine equivalent workflows.\n\n| Purpose | Capability |\n|---|---|\n| `procedure-skill-improvement` | Improve reusable procedures through bounded candidates, held-out evaluation, authorized promotion, and rollback |\n| `ideation` | Generate and refine candidate approaches before commitment |\n| `planning-decomposition` | Turn an accepted outcome into an ordered plan and bounded work items |\n| `implementation-execution` | Execute an accepted plan incrementally without changing its governing contract |\n| `test-driven-development` | Drive permanent behavior changes through a discriminating failing check and red–green–refactor |\n| `debugging-remediation` | Reproduce, isolate, repair, and verify a concrete defect |\n| `completion-verification` | Challenge completion claims with evidence from the exact candidate revision |\n| `generic-code-review` | Review a candidate for correctness, quality, and actionable defects |\n| `bounded-requirements-specification` | Write a bounded, implementation-ready specification for an accepted change |\n\n#### Provider capability families\n\nEvery family has an included bounded adaptation or mapped handbook fallback.\nCompatible upstream originals are optional, trust-checked alternatives—not\nbundled upstream runtimes, installers, hooks, or an endorsement of every feature\nin the upstream project.\n\n| Family | Included capability coverage | Original mappings |\n|---|---|---|\n| [Superpowers inner loop](handbook/software-engineering/references.md) | Brainstorming, isolated workspaces, planning and execution, bounded agent coordination, TDD, debugging, verification, review requests and feedback, branch completion, and procedure improvement | 13 |\n| [Ponytail simplicity](skills/software-engineering-handbook/experts/ponytail-simplicity.md) | Comprehension-first minimal implementation, over-engineering review, broader simplicity audit, and shortcut-debt reporting | 4 |\n| [Agent Skills: requirements and planning](skills/software-engineering-handbook/experts/requirements-planning.md) | Consequential clarification, idea refinement, specification, constraint-first design, and task decomposition | 5 |\n| [Agent Skills: context and sources](skills/software-engineering-handbook/experts/context-sources.md) | Focused context acquisition, primary-source grounding, and doubt-driven challenge | 3 |\n| [Agent Skills: interfaces and UI](skills/software-engineering-handbook/experts/interfaces-ui.md) | Frontend and user-facing behavior, plus public API and module contracts | 2 |\n| [Agent Skills: quality and operations](skills/software-engineering-handbook/experts/quality-operations.md) | Incremental implementation, TDD, browser verification, debugging, review, simplification, security, performance, Git/versioning, CI/CD, migrations, documentation/ADRs, observability, and launch | 14 |\n| [Understand Anything](skills/software-engineering-handbook/experts/code-comprehension.md) | Task-scoped codebase mapping, architecture orientation, diff impact, explanations, and onboarding | 4 |\n| [autoresearch](skills/software-engineering-handbook/experts/empirical-optimization.md) | Finite empirical optimization under an owner-approved frozen protocol | 1 |\n| [Spec Kit](skills/software-engineering-handbook/experts/spec-delivery.md) | Durable specification chains, plans/tasks, issue conversion, implementation convergence, clarification, consistency analysis, checklists, bug assessment/repair/tests, and staged opportunity assessment | 18 |\n| [OpenSpec](skills/software-engineering-handbook/experts/spec-delivery.md) | Brownfield exploration, proposals and change packets, staged or fast-forward artifact preparation, implementation, verification, provenance-retaining archive, and onboarding | 10 |\n| [BMAD Method](skills/software-engineering-handbook/experts/adaptive-agile.md) | Project context, bounded build/spec flows, course correction, brainstorming, product briefs/PRFAQs/PRDs, UX, architecture, epics/stories, sprint planning, review, checkpoint previews, end-to-end test design, and retrospectives | 16 |\n\n<details>\n<summary>All 90 registered original identifiers, grouped by capability family</summary>\n\nThese are the complete registered mappings, including aliases belonging to the\nshared purposes above. Consult the registry and source register for exact\ntriggers, exclusions, source identities, effects, and fallback boundaries.\n\n**Superpowers inner loop (13)**\n\n- `superpowers/brainstorming`\n- `superpowers/using-git-worktrees`\n- `superpowers/writing-plans`\n- `superpowers/executing-plans`\n- `superpowers/subagent-driven-development`\n- `superpowers/dispatching-parallel-agents`\n- `superpowers/test-driven-development`\n- `superpowers/systematic-debugging`\n- `superpowers/verification-before-completion`\n- `superpowers/requesting-code-review`\n- `superpowers/receiving-code-review`\n- `superpowers/finishing-a-development-branch`\n- `superpowers/writing-skills`\n\n**Ponytail simplicity (4)**\n\n- `ponytail/ponytail`\n- `ponytail/ponytail-review`\n- `ponytail/ponytail-audit`\n- `ponytail/ponytail-debt`\n\n**Agent Skills: requirements and planning (5)**\n\n- `agent-skills/interview-me`\n- `agent-skills/idea-refine`\n- `agent-skills/spec-driven-development`\n- `agent-skills/constraint-driven-development`\n- `agent-skills/planning-and-task-breakdown`\n\n**Agent Skills: context and sources (3)**\n\n- `agent-skills/context-engineering`\n- `agent-skills/source-driven-development`\n- `agent-skills/doubt-driven-development`\n\n**Agent Skills: interfaces and UI (2)**\n\n- `agent-skills/frontend-ui-engineering`\n- `agent-skills/api-and-interface-design`\n\n**Agent Skills: quality and operations (14)**\n\n- `agent-skills/incremental-implementation`\n- `agent-skills/test-driven-development`\n- `agent-skills/browser-testing-with-devtools`\n- `agent-skills/debugging-and-error-recovery`\n- `agent-skills/code-review-and-quality`\n- `agent-skills/code-simplification`\n- `agent-skills/security-and-hardening`\n- `agent-skills/performance-optimization`\n- `agent-skills/git-workflow-and-versioning`\n- `agent-skills/ci-cd-and-automation`\n- `agent-skills/deprecation-and-migration`\n- `agent-skills/documentation-and-adrs`\n- `agent-skills/observability-and-instrumentation`\n- `agent-skills/shipping-and-launch`\n\n**Understand Anything (4)**\n\n- `understand-anything/understand`\n- `understand-anything/understand-diff`\n- `understand-anything/understand-explain`\n- `understand-anything/understand-onboard`\n\n**autoresearch (1)**\n\n- `autoresearch/autoresearch`\n\n**Spec Kit (18)**\n\n- `spec-kit/speckit-constitution`\n- `spec-kit/speckit-specify`\n- `spec-kit/speckit-plan`\n- `spec-kit/speckit-tasks`\n- `spec-kit/speckit-taskstoissues`\n- `spec-kit/speckit-implement`\n- `spec-kit/speckit-converge`\n- `spec-kit/speckit-clarify`\n- `spec-kit/speckit-analyze`\n- `spec-kit/speckit-checklist`\n- `spec-kit/speckit-bug-assess`\n- `spec-kit/speckit-bug-fix`\n- `spec-kit/speckit-bug-test`\n- `spec-kit/speckit-assess-intake`\n- `spec-kit/speckit-assess-research`\n- `spec-kit/speckit-assess-define`\n- `spec-kit/speckit-assess-shape`\n- `spec-kit/speckit-assess-decide`\n\n**OpenSpec (10)**\n\n- `openspec/openspec-explore`\n- `openspec/openspec-propose`\n- `openspec/openspec-new`\n- `openspec/openspec-continue`\n- `openspec/openspec-ff`\n- `openspec/openspec-apply`\n- `openspec/openspec-verify`\n- `openspec/openspec-archive`\n- `openspec/openspec-bulk-archive`\n- `openspec/openspec-onboard`\n\n**BMAD Method (16)**\n\n- `bmad-method/bmad-project-context`\n- `bmad-method/bmad-build`\n- `bmad-method/bmad-spec`\n- `bmad-method/bmad-correct-course`\n- `bmad-method/bmad-brainstorming`\n- `bmad-method/bmad-product-brief`\n- `bmad-method/bmad-prfaq`\n- `bmad-method/bmad-prd`\n- `bmad-method/bmad-ux`\n- `bmad-method/bmad-architecture`\n- `bmad-method/bmad-create-epics-and-stories`\n- `bmad-method/bmad-sprint-planning`\n- `bmad-method/bmad-code-review`\n- `bmad-method/bmad-checkpoint-preview`\n- `bmad-method/bmad-qa-generate-e2e-tests`\n- `bmad-method/bmad-retrospective`\n\n</details>\n\n### Routing and package utilities\n\n- **Sparse selection:** risk-scaled expert budgets, one primary workflow expert,\n  and at most one original or fallback for each purpose.\n- **Complete route contracts:** triggers and near-miss exclusions, prerequisites,\n  required dependency closure, conflicts and ordering, source/trust identity,\n  permitted effects, evidence, failure behavior, and rollback.\n- **Bounded descriptor lookup:** an optional standard-library Python CLI accepts\n  a purpose or provider-qualified original, works from any directory, and\n  reports invalid inputs explicitly without selecting or activating a provider.\n- **Catalog maintenance:** included validation utilities cover registry/schema\n  consistency, mapping completeness, sparse routing, context budgets, pressure,\n  effects, and failover; behavioral checks cover the descriptor query.\n- **Portable packaging:** [11 documented harness surfaces](#install-on-supported-agents),\n  small discovery adapters, complete internal fallbacks, immutable-ref\n  installation where supported, and source/version verification guidance.\n- **Explicit limits:** no bundled upstream provider runtimes, hidden telemetry,\n  autonomous publishing/deployment, universal fixed test-coverage threshold, or\n  unmeasured quality, token-cost, or performance guarantee. Source-specific\n  exclusions remain in the [source register](handbook/software-engineering/references.md).\n\n## How it works without filling the context window\n\nThe package behaves like a book with an index instead of putting every page in\nworking memory.\n\n| Layer | Loaded when | Content |\n|---|---|---|\n| Discovery | Agent startup | Skill name and description only; capped at 1 KiB |\n| Router index | A matching engineering task begins | Compact `SKILL.md`; capped at 10 KiB |\n| Working set | The router resolves the task | Only applicable handbook and expert sections |\n| Cold storage | Until selected | Full chapters, registry, sources, and inactive experts |\n\nCatalog growth does not expand the active prompt. The adapters in this\nrepository register skill paths but do not inject the full handbook at startup.\nAfter compaction, agents retain concise decisions and evidence, then re-read\nonly authoritative sections that are still needed.\n\nFor ambiguous discovery on hosts that support command execution, an optional\nPython-standard-library query prints one purpose's complete descriptors and\npolicy instead of both cold catalogs:\n\n```sh\npython3 managed-skills/software-engineering-handbook/query_experts.py ideation\n```\n\nThe script also accepts a provider-qualified original, works by absolute path\nfrom any directory, preserves ordered alternatives and required dependencies,\nand never loads bodies, selects a route, infers trust, or activates a provider.\nIt is not an additional ordinary-task preflight. See the\n[resolution contract](skills/software-engineering-handbook/experts/resolution.md)\nfor unavailable-tool behavior.\n\n## Directory roles\n\nThey are related, but not duplicates:\n\n| Path | Role |\n|---|---|\n| `handbook/software-engineering/` | The normative, human-readable book—the long-term knowledge layer |\n| `skills/software-engineering-handbook/` | The canonical portable router, expert modules, and registry |\n| `managed-skills/software-engineering-handbook/` | Package query/validation utilities and the tiny OMP adapter |\n| `rules/engineering-handbook-enforcement.md` | OMP-specific global enforcement adapter |\n\n`software-engineering` names the handbook subject. `software-engineering-handbook`\nis the installable skill/package identity. Keeping the book separate from the\nrouter lets humans browse it normally while agents load only selected sections.\n\n## Install on supported agents\n\nInstallation changes local agent configuration and packages can influence agent\nactions. Review the repository and pin a trusted tag or commit when your harness\nsupports refs.\n\n### Claude App\n\nOpen **Customize → Plugins → Personal plugins → + → Add marketplace**, add:\n\n```text\nhttps://github.com/krunaldodiya/software-engineering-handbook\n```\n\nInstall **Software Engineering Handbook**. The bundled skill works in Claude\nweb chat, Claude Desktop chat, and Cowork.\n\n### Claude Code\n\n```text\n/plugin marketplace add krunaldodiya/software-engineering-handbook\n/plugin install software-engineering-handbook@software-engineering-handbook\n```\n\nStart a new Claude Code session.\n\n### Antigravity\n\n```sh\nagy plugin install https://github.com/krunaldodiya/software-engineering-handbook\n```\n\n### Codex App\n\nOpen **Settings → Plugins → Add Marketplace**, add:\n\n```text\nhttps://github.com/krunaldodiya/software-engineering-handbook\n```\n\nInstall **Software Engineering Handbook** from that marketplace and start a new\nchat.\n\n### Codex CLI\n\n```sh\ncodex plugin marketplace add krunaldodiya/software-engineering-handbook\ncodex plugin add software-engineering-handbook@software-engineering-handbook\n```\n\n### Gemini CLI\n\n```sh\ngemini extensions install https://github.com/krunaldodiya/software-engineering-handbook\n```\n\nGemini loads only the small `GEMINI.md` routing pointer at startup.\n\n### Kimi Code CLI\n\n```text\n/plugins install https://github.com/krunaldodiya/software-engineering-handbook\n```\n\nRun `/reload` or start a new Kimi session after installation.\n\n### OpenCode\n\nAdd the Git-backed package to the `plugin` array in global or project\n`opencode.json`:\n\n```json\n{\n  \"plugin\": [\n    \"software-engineering-handbook@git+https://github.com/krunaldodiya/software-engineering-handbook.git\"\n  ]\n}\n```\n\nRestart OpenCode. Its small adapter registers the skill path without injecting\nhandbook content.\n\n### Hermes Agent\n\n```sh\nhermes plugins install krunaldodiya/software-engineering-handbook --enable\n```\n\nHermes uses the code-free Agent Plugins v1 manifest at `plugin.json`. It\ndiscovers the portable skill without a startup hook or router-body injection.\n\n### Pi\n\n```sh\npi install git:github.com/krunaldodiya/software-engineering-handbook\n```\n\nTry without installing:\n\n```sh\npi -e git:github.com/krunaldodiya/software-engineering-handbook\n```\n\n### Oh My Pi\n\n```sh\nomp plugin install github:krunaldodiya/software-engineering-handbook\n```\n\nStart a new session after installation. Invoke\n`/skill:software-engineering-handbook` explicitly where skill commands are\nsupported.\n\nTo update or reinstall a reviewed revision, rerun `omp plugin install` with\n`github:krunaldodiya/software-engineering-handbook#<reviewed-commit-sha>`.\nUninstalling first is unnecessary. Verify `omp plugin list --json` and the\ninstalled package identity, then have a fresh OMP session read\n`skill://software-engineering-handbook`; a native skill can shadow a plugin\nskill. Keep the prior reviewed ref for rollback. Reinstallation changes future\ndiscovery, not the instructions already loaded into an active conversation.\n\n## Future harness support\n\nThese integrations are intentionally deferred:\n\n- [ ] Cursor\n- [ ] Devin CLI\n- [ ] Factory Droid\n- [ ] GitHub Copilot CLI\n- [ ] Grok Build CLI\n\n## Updating\n\nUse the selected harness's normal plugin or skill update mechanism. For\nreproducible use, install a reviewed tag or commit instead of a moving branch\nwhen the harness supports immutable references.\n\n## Repository contents\n\n- `handbook/software-engineering/` — portable normative handbook and source\n  register.\n- `skills/software-engineering-handbook/` — canonical portable router, expert\n  modules, and registry.\n- `managed-skills/software-engineering-handbook/` — package query/validation\n  utilities and the tiny OMP adapter.\n- Harness manifests under `.agents/`, `.claude-plugin/`, `.codex-plugin/`,\n  `.kimi-plugin/`, and `.opencode/`, plus root Agent Plugins v1, Gemini, and Pi\n  manifests.\n- `rules/engineering-handbook-enforcement.md` — OMP global-rule adapter.\n\n## Validate and contribute\n\nValidate the catalog and context budgets before proposing a change:\n\n```sh\npython3 managed-skills/software-engineering-handbook/validate_registry.py\npython3 -O managed-skills/software-engineering-handbook/validate_registry.py\npython3 managed-skills/software-engineering-handbook/test_query_experts.py\npython3 -O managed-skills/software-engineering-handbook/test_query_experts.py\n```\n\nFor a provider capability change:\n\n1. Pin and review the exact upstream source.\n2. Map every original to an existing purpose in `experts/purposes.json`. Add a\n   new purpose only for a demonstrably distinct behavioral boundary.\n3. Gap-map it against existing experts.\n4. Reuse an existing module or add one bounded expert module.\n5. Register triggers, exclusions, prerequisites, conflicts, effects, evidence,\n   context budget, and rollback.\n6. Preserve ordered original preference and the single purpose fallback.\n7. Add pressure checks for classification, duplicate routes, presence, absence,\n   conflict, and active-budget paths.\n8. Obtain exact-revision review before publication.\n\nDo not add another workflow router, whole-catalog loading, hidden installation,\ntelemetry, or a capability already covered by a compatible expert.\n\nIssues and pull requests are welcome. Keep changes evidence-backed, portable,\nand narrowly owned.\n\n## Safety and repository boundary\n\nThis public repository contains only handbook and packaging assets. Never commit\nagent configuration, environment files, databases, caches, sessions,\ncredentials, secrets, private project adapters, or unrelated files.\n\nAgent plugins and skills can influence actions. Installation does not authorize\npublishing, deployment, destructive operations, secret access, or any other\nprotected effect. Normal user, project, and harness approvals still apply.\n\n## Upstream attribution\n\nThis repository independently adapts workflow concepts; it does not vendor\nupstream skill bodies or runtimes. Reviewed upstream revisions, licenses,\nexclusions, and provenance links are recorded in the\n[source register](handbook/software-engineering/references.md). Project names\nand trademarks remain property of their respective owners.\n\n## License\n\n[MIT](LICENSE)\n",
  "bytes": 26331,
  "sha": "deb7bc955e3e811a9f38c97eda3e6aab669aacdd8bac41904cae87b9569b2d5d",
  "repo_slug": "krunaldodiya/software-engineering-handbook",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_krunaldodiya_software_engineering_handbo_a744bd5a/readme"
}