seedproject-web/agents/agents.json

265 lines
7.5 KiB
JSON
Raw Normal View History

feat: content-pipeline/ → agents/ — formalize the agent system in the seed Adopt the agents/ architecture proven on medellin.co (reference impl): - Move the content engine to a top-level agents/ dir: orchestrators, prompts, config, run.sh, admin console, shared libs. All content-pipeline literals repointed (config paths, scripts, admin, LLM-facing prompts/image.md string, configure.mjs, new-site.sh, astroagent tokenFile, .gitignore runtime block). - Every script carries a parseable @agent-manifest header: name, title, class (content|operational|runtime|plumbing), trigger, model, prompts, skills (MCP), tools, reads/writes tables. 5 content agents + 3 plumbing scripts. - New agents/catalog.mjs generates the catalog from the headers: agents/AGENTS.md (human, grouped by class) + agents/agents.json (machine manifest — a clone diffs it against a source to find missing tools/tables/MCP before running). configure.mjs regenerates the catalog on every identity stamp. No DB table, no watcher. - config.json gains paths.stateDir/newsDir; publish-tick, write-daily, and news-radar read them instead of hardcoding. - Full cut: content-pipeline/ deleted (the seed has no live crons, so no hybrid period needed). Docs updated (AGENTS.md structure + pipeline section, README paths). Clones migrating from content-pipeline/: see medellin.co's .memory/handoffs/agents-directory-migration.md for the cutover playbook (one cron set active at a time; migrate drafts/state after repointing cron). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FMQeUnUrAeexcZ7P2Hxa6G
2026-07-11 20:33:48 +00:00
{
"agents": [
{
"file": "agents/scripts/approve.mjs",
"name": "approve",
"title": "Draft Approver",
"class": "plumbing",
"trigger": "manual <slug> [--force]",
"description": "Approve a draft: SEO gate, then promote it into the blog, build, go live. No LLM.",
"model": "-",
"prompt": [],
"skills": [],
"tools": [
"lib/publish.mjs"
],
"reads": [],
"writes": [],
"created": "2026-07-04"
},
{
"file": "agents/console/agents/in-page-console.mjs",
"name": "in-page-console",
"title": "In-page Console",
"class": "runtime",
"trigger": "POST /run; GET /stream; POST /publish; POST /rebuild; POST /discard",
"model": "claude-sonnet-4-6",
"description": "Freeform ▲ editor: describe a change on any page; the agent edits, the console builds an isolated preview, then publish/discard.",
"prompt": [],
"skills": [],
"tools": [
"Read",
"Write",
"Edit",
"Glob",
"Grep",
"WebSearch"
],
"reads": [],
"writes": [],
"created": ""
},
feat: content-pipeline/ → agents/ — formalize the agent system in the seed Adopt the agents/ architecture proven on medellin.co (reference impl): - Move the content engine to a top-level agents/ dir: orchestrators, prompts, config, run.sh, admin console, shared libs. All content-pipeline literals repointed (config paths, scripts, admin, LLM-facing prompts/image.md string, configure.mjs, new-site.sh, astroagent tokenFile, .gitignore runtime block). - Every script carries a parseable @agent-manifest header: name, title, class (content|operational|runtime|plumbing), trigger, model, prompts, skills (MCP), tools, reads/writes tables. 5 content agents + 3 plumbing scripts. - New agents/catalog.mjs generates the catalog from the headers: agents/AGENTS.md (human, grouped by class) + agents/agents.json (machine manifest — a clone diffs it against a source to find missing tools/tables/MCP before running). configure.mjs regenerates the catalog on every identity stamp. No DB table, no watcher. - config.json gains paths.stateDir/newsDir; publish-tick, write-daily, and news-radar read them instead of hardcoding. - Full cut: content-pipeline/ deleted (the seed has no live crons, so no hybrid period needed). Docs updated (AGENTS.md structure + pipeline section, README paths). Clones migrating from content-pipeline/: see medellin.co's .memory/handoffs/agents-directory-migration.md for the cutover playbook (one cron set active at a time; migrate drafts/state after repointing cron). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FMQeUnUrAeexcZ7P2Hxa6G
2026-07-11 20:33:48 +00:00
{
"file": "agents/scripts/list-drafts.mjs",
"name": "list-drafts",
"title": "Draft Lister",
"class": "plumbing",
"trigger": "manual",
"description": "List pending drafts awaiting review. No LLM.",
"model": "-",
"prompt": [],
"skills": [],
"tools": [],
"reads": [],
"writes": [],
"created": "2026-07-04"
},
{
"file": "agents/scripts/news-radar.mjs",
"name": "news-radar",
"title": "News Radar",
"class": "content",
"trigger": "manual / cron-capable (maintains news-queue.json for the writer)",
"description": "Discover timely niche news and maintain news-queue.json; the daily writer drains it first.",
"model": "research=claude-opus-4-8",
"prompt": [
"prompts/news-radar.system.md"
],
"skills": [],
"tools": [
"lib/claude.mjs"
],
"reads": [],
"writes": [],
"created": "2026-07-04"
},
{
"file": "agents/console/agents/project-builder.mjs",
"name": "project-builder",
"title": "Project Builder",
"class": "runtime",
"trigger": "POST /draft; POST /build-project",
"model": "claude-sonnet-4-6",
"description": "Drafts copy and authors a full structured project record (reads uploaded images) for the New Project admin form. Returns JSON; never mutates the repo.",
"prompt": [],
"skills": [],
"tools": [
"Read"
],
"reads": [
"cja_projects"
],
"writes": [
"cja_projects"
],
"created": ""
},
feat: content-pipeline/ → agents/ — formalize the agent system in the seed Adopt the agents/ architecture proven on medellin.co (reference impl): - Move the content engine to a top-level agents/ dir: orchestrators, prompts, config, run.sh, admin console, shared libs. All content-pipeline literals repointed (config paths, scripts, admin, LLM-facing prompts/image.md string, configure.mjs, new-site.sh, astroagent tokenFile, .gitignore runtime block). - Every script carries a parseable @agent-manifest header: name, title, class (content|operational|runtime|plumbing), trigger, model, prompts, skills (MCP), tools, reads/writes tables. 5 content agents + 3 plumbing scripts. - New agents/catalog.mjs generates the catalog from the headers: agents/AGENTS.md (human, grouped by class) + agents/agents.json (machine manifest — a clone diffs it against a source to find missing tools/tables/MCP before running). configure.mjs regenerates the catalog on every identity stamp. No DB table, no watcher. - config.json gains paths.stateDir/newsDir; publish-tick, write-daily, and news-radar read them instead of hardcoding. - Full cut: content-pipeline/ deleted (the seed has no live crons, so no hybrid period needed). Docs updated (AGENTS.md structure + pipeline section, README paths). Clones migrating from content-pipeline/: see medellin.co's .memory/handoffs/agents-directory-migration.md for the cutover playbook (one cron set active at a time; migrate drafts/state after repointing cron). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FMQeUnUrAeexcZ7P2Hxa6G
2026-07-11 20:33:48 +00:00
{
"file": "agents/scripts/publish-tick.mjs",
"name": "publish-tick",
"title": "Auto-Publisher Tick",
"class": "plumbing",
"trigger": "cron:run.sh publish-tick.mjs(15m)",
"description": "Once per day after the morning floor, publish ONE eligible draft (SEO-gated) with a randomized earlier-today timestamp. No LLM.",
"model": "-",
"prompt": [],
"skills": [],
"tools": [
"lib/publish.mjs"
],
"reads": [],
"writes": [],
"created": "2026-07-04"
},
{
"file": "agents/console/agents/qa.mjs",
"name": "qa",
"title": "QA Agent",
"class": "runtime",
"trigger": "POST /qa/run; schedule(heartbeatMin)",
"model": "claude-sonnet-4-6",
"description": "Crawls the live site (links, images, forms, API, SEO/meta, a11y); auto-fixes safe findings via the Web Designer queue; runs on a heartbeat.",
"prompt": [],
"skills": [
"qa"
],
"tools": [
"Read",
"Skill",
"api/cli/qa-start.php",
"api/cli/qa-finish.php",
"api/cli/qa-routes.php",
"api/cli/qa-autofix.php"
],
"reads": [
"cja_qa_runs",
"cja_qa_findings"
],
"writes": [
"cja_qa_runs",
"cja_qa_findings"
],
"created": ""
},
feat: content-pipeline/ → agents/ — formalize the agent system in the seed Adopt the agents/ architecture proven on medellin.co (reference impl): - Move the content engine to a top-level agents/ dir: orchestrators, prompts, config, run.sh, admin console, shared libs. All content-pipeline literals repointed (config paths, scripts, admin, LLM-facing prompts/image.md string, configure.mjs, new-site.sh, astroagent tokenFile, .gitignore runtime block). - Every script carries a parseable @agent-manifest header: name, title, class (content|operational|runtime|plumbing), trigger, model, prompts, skills (MCP), tools, reads/writes tables. 5 content agents + 3 plumbing scripts. - New agents/catalog.mjs generates the catalog from the headers: agents/AGENTS.md (human, grouped by class) + agents/agents.json (machine manifest — a clone diffs it against a source to find missing tools/tables/MCP before running). configure.mjs regenerates the catalog on every identity stamp. No DB table, no watcher. - config.json gains paths.stateDir/newsDir; publish-tick, write-daily, and news-radar read them instead of hardcoding. - Full cut: content-pipeline/ deleted (the seed has no live crons, so no hybrid period needed). Docs updated (AGENTS.md structure + pipeline section, README paths). Clones migrating from content-pipeline/: see medellin.co's .memory/handoffs/agents-directory-migration.md for the cutover playbook (one cron set active at a time; migrate drafts/state after repointing cron). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FMQeUnUrAeexcZ7P2Hxa6G
2026-07-11 20:33:48 +00:00
{
"file": "agents/scripts/research.mjs",
"name": "research",
"title": "Editorial Calendar Researcher",
"class": "content",
"trigger": "manual (auto-invoked by writer when the evergreen backlog runs low)",
"description": "Top up the evergreen backlog in calendar.json (additive, deduped); timely news is the news radar's job.",
"model": "research=claude-opus-4-8",
"prompt": [
"prompts/research.system.md"
],
"skills": [],
"tools": [
"lib/claude.mjs",
"lib/calendar.mjs"
],
"reads": [],
"writes": [],
"created": "2026-07-04"
},
{
"file": "agents/scripts/revise.mjs",
"name": "reviser",
"title": "Draft Reviser",
"class": "content",
"trigger": "manual <slug> (auto-invoked when the SEO gate fails and autoRevise is on)",
"description": "Feed the SEO audit's fixes back to the writer and re-audit until the draft passes or attempts run out.",
"model": "writer=claude-sonnet-4-6",
"prompt": [
"prompts/reviser.system.md"
],
"skills": [],
"tools": [
"lib/claude.mjs"
],
"reads": [],
"writes": [],
"created": "2026-07-04"
},
{
"file": "agents/scripts/seo-review.mjs",
"name": "seo-review",
"title": "SEO Reviewer",
"class": "content",
"trigger": "manual <slug> (auto-invoked by writer/approve gates)",
"description": "Audit one article for EEAT / spam-policy / on-page SEO / AEO / readability → seo-review.json.",
"model": "reviewer=claude-opus-4-8",
"prompt": [
"prompts/seo-review.system.md"
],
"skills": [],
"tools": [
"lib/claude.mjs"
],
"reads": [],
"writes": [],
"created": "2026-07-04"
},
{
"file": "agents/console/agents/web-designer.mjs",
"name": "web-designer",
"title": "Web Designer",
"class": "runtime",
"trigger": "POST /tasks/run",
"model": "claude-sonnet-4-6",
"description": "Durable design queue (cja_tasks): runs the full-builder agent on any page from a brief, auto-publishes (build-gated scoped commit), and self-logs to the changelog. Fed by the admin and by QA autofix.",
"prompt": [],
"skills": [
"brand",
"ui-ux",
"changelog"
],
"tools": [
"Read",
"Write",
"Edit",
"Glob",
"Grep",
"WebSearch",
"Skill",
"api/cli/tasks-next.php",
"api/cli/tasks-finish.php",
"api/cli/seed-changelog.php"
],
"reads": [
"cja_tasks",
"cja_changelog",
"cja_projects"
],
"writes": [
"cja_tasks",
"cja_changelog",
"cja_projects"
],
"created": ""
},
feat: content-pipeline/ → agents/ — formalize the agent system in the seed Adopt the agents/ architecture proven on medellin.co (reference impl): - Move the content engine to a top-level agents/ dir: orchestrators, prompts, config, run.sh, admin console, shared libs. All content-pipeline literals repointed (config paths, scripts, admin, LLM-facing prompts/image.md string, configure.mjs, new-site.sh, astroagent tokenFile, .gitignore runtime block). - Every script carries a parseable @agent-manifest header: name, title, class (content|operational|runtime|plumbing), trigger, model, prompts, skills (MCP), tools, reads/writes tables. 5 content agents + 3 plumbing scripts. - New agents/catalog.mjs generates the catalog from the headers: agents/AGENTS.md (human, grouped by class) + agents/agents.json (machine manifest — a clone diffs it against a source to find missing tools/tables/MCP before running). configure.mjs regenerates the catalog on every identity stamp. No DB table, no watcher. - config.json gains paths.stateDir/newsDir; publish-tick, write-daily, and news-radar read them instead of hardcoding. - Full cut: content-pipeline/ deleted (the seed has no live crons, so no hybrid period needed). Docs updated (AGENTS.md structure + pipeline section, README paths). Clones migrating from content-pipeline/: see medellin.co's .memory/handoffs/agents-directory-migration.md for the cutover playbook (one cron set active at a time; migrate drafts/state after repointing cron). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FMQeUnUrAeexcZ7P2Hxa6G
2026-07-11 20:33:48 +00:00
{
"file": "agents/scripts/write-daily.mjs",
"name": "writer",
"title": "Daily Blog Writer",
"class": "content",
"trigger": "cron:run.sh write-daily.mjs(daily)",
"description": "Draft one article (MDX + cover + sources) — today's news first, else next evergreen calendar topic.",
"model": "writer=claude-sonnet-4-6",
"prompt": [
"prompts/writer.system.md",
"prompts/image.md"
],
"skills": [
"mcp__claude_ai_Higgsfield"
],
"tools": [
"lib/claude.mjs",
"lib/calendar.mjs"
],
"reads": [],
"writes": [],
"created": "2026-07-04"
}
]
}