fimo 2.10.1-staging.1 → 2.10.1-staging.3

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (43) hide show
  1. package/assets/skills/fimo-cli/SKILL.md +57 -139
  2. package/assets/skills/fimo-cli/SKILL.studio.md +42 -0
  3. package/assets/skills/fimo-cli/references/branches.md +43 -106
  4. package/assets/skills/fimo-cli/references/content/bulk.md +48 -0
  5. package/assets/skills/fimo-cli/references/content/entries.md +54 -0
  6. package/assets/skills/fimo-cli/references/content/localized.md +56 -0
  7. package/assets/skills/fimo-cli/references/content/schemas.md +29 -0
  8. package/assets/skills/fimo-cli/references/content/sql.md +57 -0
  9. package/assets/skills/fimo-cli/references/content.md +33 -272
  10. package/assets/skills/fimo-cli/references/deploy.md +35 -126
  11. package/assets/skills/fimo-cli/references/domains.md +38 -131
  12. package/assets/skills/fimo-cli/references/forms.md +1 -1
  13. package/assets/skills/fimo-cli/references/init.md +36 -170
  14. package/assets/skills/fimo-cli/references/locales.md +2 -2
  15. package/assets/skills/fimo-cli/references/projects.md +1 -1
  16. package/assets/skills/fimo-code/references/design.md +42 -126
  17. package/assets/skills/fimo-migration/SKILL.md +114 -408
  18. package/assets/skills/fimo-migration/references/content-delivery.md +67 -0
  19. package/assets/skills/fimo-migration/references/frameworks/nextjs.md +1 -1
  20. package/assets/skills/fimo-migration/references/qa/browser.md +41 -0
  21. package/assets/skills/fimo-migration/references/qa/guardrails.md +95 -0
  22. package/assets/skills/fimo-migration/references/qa/results.md +78 -0
  23. package/assets/skills/fimo-migration/references/qa.md +102 -383
  24. package/assets/skills/fimo-migration/references/questions-and-improvements.md +77 -168
  25. package/assets/skills/fimo-migration/references/setup.md +90 -0
  26. package/assets/skills/fimo-migration/references/workers/validator.md +2 -2
  27. package/assets/skills/fimo-migration/references/workers.md +1 -1
  28. package/assets/skills/fimo-migration/references/workflow.md +117 -171
  29. package/assets/skills/fimo-studio/SKILL.md +12 -11
  30. package/assets/skills/fimo-studio/references/ai-chat/agents.md +46 -0
  31. package/assets/skills/fimo-studio/references/ai-chat/default-agents.md +57 -0
  32. package/assets/skills/fimo-studio/references/ai-chat/routines.md +53 -0
  33. package/assets/skills/fimo-studio/references/ai-chat.md +39 -76
  34. package/assets/skills/fimo-studio/references/custom-domains.md +2 -2
  35. package/assets/skills/fimo-studio/references/publish-and-deploy.md +1 -1
  36. package/dist/cli/bundle.json +1 -1
  37. package/dist/cli/index.js +38 -17
  38. package/package.json +8 -8
  39. package/templates/astro/package.json +3 -3
  40. package/templates/nextjs/package.json +4 -4
  41. package/templates/react-router/package.json +5 -5
  42. package/templates/sveltekit/package.json +3 -3
  43. package/templates/vite/package.json +5 -5
@@ -1,184 +1,130 @@
1
- # Migration Workflow
2
-
3
- Use this as the phase contract for a complete migration. The coordinator and
4
- workers communicate in plain text. JSON is used only for Fimo schemas and the
5
- content payloads that the data-plane commands import.
6
-
7
- ## Preflight: inspect and ask
8
-
9
- Before target initialization or any other write, inspect the source read-only
10
- and follow `references/questions-and-improvements.md`. Use the coding agent's
11
- native structured-question UI to ask one stacked batch for integration,
12
- environment, framework, and ownership decisions that the source cannot settle.
13
- Never move environment values or credentials without explicit consent. Do not
14
- ask a generic optimization question.
15
-
16
- ## Phase 0: Initialize the target
17
-
18
- 1. Pick the setup row in SKILL.md "Initialized target precondition", run it,
19
- and pass its gate before workers start. Framework and `--source-platform`
20
- rules for `fimo migrate init` live in SKILL.md "Target setup".
21
- 2. Inspect the target's real `package.json`, `.fimo/config.json`, framework config, and route
22
- directories. They are the authority for paths. Do not infer an app root
23
- from a skill example and do not create a second `app/` or `src/app/` tree.
24
- 3. Verify dependency provenance before workers start. The target pins the
25
- CLI artifact that set it up: an exact published release with matching
26
- companion packages, or the exact dev artifact from SKILL.md "CLI
27
- provenance". Stop when the target records an absolute package-manager cache
28
- path, a `file:` link, an unapproved `latest`, or a CLI and API from
29
- different runtime channels; do not let the producer repair release setup
30
- inside the migration.
1
+ # Migration workflow
2
+
3
+ The phases of a complete migration. The coordinator and workers talk in plain
4
+ text. JSON is only for Fimo schemas, import payloads, and QA guardrail
5
+ evidence.
6
+
7
+ ## Preflight
8
+
9
+ Inspect the source read-only and ask one question batch
10
+ (`questions-and-improvements`) before any write.
11
+
12
+ ## Phase 0: Set up the target
13
+
14
+ 1. Run the case from SKILL.md "Initialized target precondition" (`setup` has
15
+ the commands) and pass its setup gate before workers start.
16
+ 2. Read the target's real `package.json`, `.fimo/config.json`, framework
17
+ config, and route directories. They are the authority for paths.
18
+ 3. Check dependency provenance. The target pins the CLI artifact that set it
19
+ up: an exact release with matching companion packages, or the exact
20
+ development artifact from `.fimo/dev.json`. Stop when the target records a
21
+ package-manager cache path, a `file:` link, an unapproved `latest`, or a CLI
22
+ and API from different runtime channels. Do not repair release setup inside
23
+ the migration.
31
24
  4. When `fimo migrate init` scaffolded the target, keep source material in
32
- `.fimo/migration/source` or another explicit source location. Do not copy
33
- source files over the scaffolded target root.
25
+ `.fimo/migration/source` or another explicit place. Never copy it over the
26
+ target root.
34
27
 
35
- Write a short `.fimo/migration/migration-plan.md` containing the source
36
- location, selected framework, actual target directories, worker owners, and
37
- known questions. Do not create `target.json`, `manifest.json`,
38
- `worker-plan.json`, `decision.json`, or JSON worker receipts for a new run.
28
+ Write `.fimo/migration/migration-plan.md`: source location, framework, real
29
+ target directories, worker owners, dependencies, and open questions. Do not
30
+ write `target.json`, `manifest.json`, `worker-plan.json`, `decision.json`, or
31
+ JSON worker receipts.
39
32
 
40
33
  ## Phase 1: Source intake
41
34
 
42
- Read the source directly with `rg`, `find`, `ls`, and targeted file reads.
43
- Classify the available evidence:
35
+ Read the source directly (`rg`, `find`, `ls`, file reads). Find:
44
36
 
45
- - source repository, export, downloaded HTML, or media folder;
46
- - CMS CSV/XML/JSON exports and public APIs;
47
- - forms and third-party form actions;
48
- - integrations and the environment-key names they need, never secret values;
37
+ - repositories, exports, downloaded HTML, media folders, CMS CSV/XML/JSON
38
+ exports, and public APIs;
39
+ - forms and third-party form actions; integrations and their environment key
40
+ names, never values;
49
41
  - redirects, security headers, locales, SEO, robots, sitemap, and OG behavior;
50
- - unsupported behavior such as auth, commerce, search, members, custom JS, or
42
+ - unsupported behavior: auth, commerce, search, members, custom JS,
51
43
  provider-specific workers.
52
44
 
53
- For private APIs or CMSs, ask for a user-provided export or a narrowly scoped
54
- server-side credential. Never recover credentials from browser bundles or copy
55
- them into migration artifacts. If the source API URL or access is missing or
56
- unreachable, stop and ask whether to use scoped access, reconstruct only the
57
- selected published routes from the public site, or migrate code without the CMS
58
- records. Continue with public reconstruction only after that explicit choice,
59
- and record everything the rendered site cannot reveal.
60
-
61
- For API-backed content, trace active clients, endpoints, frontend types, and
62
- consumers first. Prefer backend source schemas; otherwise use public OpenAPI,
63
- GraphQL introspection, then selected REST endpoints in that order. Request a
64
- token or export only for website-required records that are not public. Retrieve
65
- all required pages, locales, relations, and media during content planning.
66
- Treat absent/private fields as unknown, keep uncertain fields optional, and do
67
- not turn sampled values into enums. Bounded retrieval scripts are accelerators
68
- the planner may adapt or replace, never schema decision makers.
69
-
70
- For URL-only migrations, read the published site yourself with a browser. Open
71
- every selected route, let client-side rendering finish, and record the sections,
72
- copy, images, links, and forms you actually see. An HTML response fetched
73
- without executing JavaScript is not evidence: a client-rendered page returns an
74
- empty root element, and treating that as the page reports an empty site as a
75
- complete one.
45
+ **Private sources.** Ask for an export or a scoped server-side credential.
46
+ Never recover credentials from browser bundles. When access is missing, ask the
47
+ missing-CMS question (`questions-and-improvements`) and wait for the choice.
48
+
49
+ **API-backed content.** First trace the active client, endpoints, frontend
50
+ types, and consumers. Then take the model from the backend source schema; else
51
+ from public OpenAPI or GraphQL introspection; else from selected REST
52
+ endpoints. Ask for a token or export only for required records that are not
53
+ public. During content planning, retrieve every page, locale, relation, and
54
+ media item, before any schema push. Treat absent or private fields as unknown,
55
+ keep uncertain fields optional, and do not make enums from sample values.
56
+ Retrieval scripts help; they never decide the schema.
57
+
58
+ **URL-only.** Read the published site with a browser. Open every selected
59
+ route, let client-side rendering finish, and record the sections, copy, images,
60
+ links, and forms you see. HTML fetched without running JavaScript is not
61
+ evidence: a client-rendered page returns an empty root, which reads as an empty
62
+ site. Never report a route as covered from a fetch that showed no sections.
63
+ For a source-backed site with a public URL (hybrid), the source is the truth
64
+ and the URL is comparison evidence.
76
65
 
77
66
  ## Phase 2: Fan out workers
78
67
 
79
- The coordinator launches only the workers supported by the source:
80
-
81
- - **runtime**: preserve the detected framework, router, runtime, dependencies,
82
- provider configuration, and environment wiring. A framework conversion needs
83
- an explicit user choice. Report active media found in this surface.
84
- - **routes/SEO**: preserve one target route per source route, redirects, 404,
85
- metadata, OG/social output, sitemap, robots, and other observable behavior.
86
- Report active media found in this surface.
87
- - **content planner**: discover content, labels, data sources, consumers, and
88
- schemas; push schemas and prepare content payloads. It does not rewrite page
89
- code or import every entry itself. Report active media found in content and
90
- API data; do not rescan unrelated runtime or route surfaces.
91
- - **locales**: configure locales and migrate translated UI-visible copy when
92
- the source has it.
93
- - **assets/forms**: after the discovery fan-in, upload the consolidated media
94
- obligations and preserve, replace, or report forms and integrations.
95
-
96
- When the user explicitly chose a framework conversion, one framework
97
- conversion worker replaces the runtime and routes lanes and owns all
98
- structural target code (see `references/workers/framework.md`). The content
99
- planner may read the source in parallel, but it must not convert the framework
100
- or rewrite page code. Follow the asset, data, and code gates in
101
- `references/workers.md`.
102
-
103
- The coordinator gives each worker a plain-language prompt naming its ownership,
104
- files, dependencies, and worker-specific reference from `references/workers/`.
105
- Workers may write the application inside their ownership. They do not need a
106
- JSON plan or receipt. If two workers need the same file, name one owner and run
107
- the other after it.
108
-
109
- ## Phase 3: Discovery fan-in and content planning
110
-
111
- The planner follows `references/workers/content-planner.md`, plus
112
- `references/content.md`, `references/decision.md`, and
113
- `references/import-contract.md`. Its text plan and payloads must be complete
114
- before Phase 4; this includes retrieving API-backed source data and tracing its
115
- consumers. Data workers must not rediscover unrelated source files.
116
-
117
- After the runtime/framework, routes/SEO, and content discovery receipts are all
68
+ Start only the lanes the source needs, as `workers` describes: runtime,
69
+ routes/SEO, content planner, and locales first. When the owner chose a
70
+ framework conversion, one framework worker replaces runtime and routes and owns
71
+ all structural code (`workers/framework`); the content planner may read the
72
+ source beside it but never converts or rewrites page code.
73
+
74
+ Give each worker a plain-language prompt with its ownership, files,
75
+ dependencies, and its `workers/` reference. If two workers need one file, name
76
+ one owner and run the other after it.
77
+
78
+ ## Phase 3: Content planning and media fan-in
79
+
80
+ The planner follows `workers/content-planner`, `content`, `decision`, and
81
+ `import-contract`. Its plan and payloads are complete before Phase 4, with
82
+ API-backed data retrieved and its consumers traced. Data workers do not
83
+ rediscover source files.
84
+
85
+ When the runtime or framework, routes/SEO, and content discovery reports are
118
86
  terminal, the coordinator combines and deduplicates their media findings into
119
- the complete media-obligation list. This is the media worker's only input.
120
-
121
- ## Phase 4: Media, forms, and integrations
122
-
123
- 1. Account for every consolidated media obligation, including schema media
124
- fields and richtext/Tiptap image nodes, plus route, SEO, and form media.
125
- 2. Upload media through Fimo and replace references only after the upload is
126
- confirmed. A local `public/` copy or source URL is not an imported asset.
127
- 3. Preserve third-party integrations unless Fimo reproduces their exact
128
- user-visible behavior. Record required environment-key names only.
129
- 4. Use a Fimo-native form only when it matches the source behavior end to end;
130
- otherwise preserve the integration or report it unresolved.
131
-
132
- After every media obligation has a target mapping, patch the content payloads
133
- and run one data worker per schema/locale. Verify each imported batch by
134
- read-back before starting its content code worker. The code worker replaces
135
- each planned consumer in place and proves at least one real imported item
136
- crosses every Server-to-Client or plain-JSON boundary it uses. A build over an
137
- empty result is not proof. Load the active `fimo-code` framework reference for
138
- the exact serialization and hydration contract.
139
-
140
- ## Phase 5: Framework implementation
141
-
142
- Preserve the detected source framework, router, runtime, and configuration.
143
- React Router/Vite stays React Router/Vite and Next.js stays Next.js. Convert to
144
- Next.js only after the user explicitly chooses it. An explicit conversion is a
145
- faithful first pass owned by the framework conversion worker: it ports the
146
- source's APIs, loaders, hardcoded content, integrations, and visible behavior
147
- into the new framework, and it must pass the framework gate in
148
- `references/workers/framework.md` before content code work continues. A gate
149
- failure is draft-only or unresolved; never fall back silently to a generic
150
- shell or another framework. Load the matching `fimo-code` framework and
151
- hosting references before integration work.
152
-
153
- Preserve the source route surface and behavior. Do not replace many pages with
154
- one generic static shell. Apply the matching framework reference before
155
- disposing of provider-specific rules; redirects, headers, feeds, metadata,
156
- forms, and freshness behavior are observable contracts.
157
-
158
- When local or bundled content becomes Fimo-backed, follow the content-delivery
159
- contract in `references/questions-and-improvements.md`. Preserve build-time or
160
- server rendering, direct lookups, bounded reads, request parallelism, and edit
161
- freshness. A new browser waterfall, repeated full-list read, avoidable
162
- sequential request chain, or stale edit is a migration regression.
163
-
164
- ## Phase 6: Fan in, validate, and report
165
-
166
- Workers return the plain-text receipt defined in `references/workers.md`.
167
- The coordinator checks those claims against the diff, source, CLI read-back,
168
- build, and rendered routes. Then it runs the read-only validation wave from
169
- `references/workers/validator.md`, passing the accepted scope and deliberate
170
- user-approved differences as task context; a worker saying "done" is not
171
- evidence.
172
- Collect the complete finding set before assigning repairs. Run disjoint repair
173
- owners together, wait for all repairs, and rerun one central validation wave.
174
-
175
- Write `.fimo/migration/migration-report.md` with route/content/media/form/SEO
176
- coverage, import counts, worker timings, unresolved items, and next actions.
177
- Use `complete` only when required checks pass, `complete-with-unresolved` when
178
- the app works and every gap is explicit, and `draft-only` when import or
179
- verification is incomplete.
180
-
181
- All checks are local. Deployment is not QA evidence. After successful local QA,
182
- run one plain `fimo deploy` for a newly created project unless the user opted out
183
- or `FIMO_MIGRATION_NO_DEPLOY=1`; for an existing linked project, offer deploy and
184
- wait for approval. Never publish automatically.
87
+ one media-obligation list. That list is the media worker's only input.
88
+
89
+ ## Phase 4: Media, forms, data, and code
90
+
91
+ 1. Account for every media obligation: schema media fields, richtext image
92
+ nodes, and route, SEO, and form media.
93
+ 2. Upload media through Fimo. Replace references only after the upload is
94
+ confirmed. A local `public/` copy or a source URL is not an imported asset.
95
+ 3. Use a Fimo form only when it matches the source behavior end to end;
96
+ otherwise keep the integration or report it unresolved.
97
+ 4. Patch the payloads with the media mappings, then run one data worker per
98
+ schema and locale. Read each imported batch back before its code worker
99
+ starts.
100
+ 5. The code worker replaces each planned consumer in place, and proves that
101
+ at least one real imported item crosses each Server-to-Client or plain-JSON
102
+ boundary it uses. A build over an empty result is not proof. Load the
103
+ `fimo-code` framework reference for serialization and hydration.
104
+
105
+ ## Phase 5: Framework and behavior
106
+
107
+ An explicit conversion is a faithful first pass: the framework worker ports the
108
+ source's APIs, loaders, hardcoded content, integrations, and visible behavior,
109
+ and passes the gate in `workers/framework` before content code continues. A
110
+ failed gate is `draft-only` or unresolved; never fall back to a generic shell
111
+ or another framework.
112
+
113
+ Load the matching `fimo-code` framework and hosting references before
114
+ integration work. Redirects, headers, feeds, metadata, forms, and freshness are
115
+ observable contracts. Classify each provider-specific file as `preserved`,
116
+ `replaced`, `safely-unnecessary-implementation-detail`, or `unresolved`, with
117
+ evidence; delete only with proof or the owner's accepted loss. Content reads
118
+ follow `content-delivery`.
119
+
120
+ ## Phase 6: Fan in, validate, repair, report
121
+
122
+ Check each worker report against the diff, source, CLI read-back, build, and
123
+ rendered routes. Then run the read-only validation wave
124
+ (`workers/validator`, `qa`) and pass it the accepted scope and the
125
+ owner-approved differences.
126
+
127
+ Collect the complete set of findings before repairs. Run repair owners with
128
+ disjoint files together, wait for all of them, then run one central validation
129
+ wave again. Write the report (`qa/results`). Deploy only as SKILL.md
130
+ "Completion and deployment" says.
@@ -1,16 +1,14 @@
1
1
  ---
2
2
  name: fimo-studio
3
3
  description: >-
4
- Load ONLY when the user explicitly asks about the Fimo web app /
5
- Fimo Studio / the dashboard UI — never for CLI commands or code work.
6
- Triggers: "in Fimo Studio", "in the dashboard", "in the web app",
7
- "how do I X in the UI", or screenshots of the dashboard. Topics
8
- covered: dashboard content and CMS, publishing from the dashboard,
9
- dashboard SEO/domain settings, site connections, schedules and run history, outbound webhooks,
10
- visual editing, version history UI, dashboard sharing/billing, dashboard chat,
11
- and the command palette. Never describe
12
- UI locations, buttons, or panels from memory — always read the relevant
13
- reference file first.
4
+ How to find and do things in the Fimo web app (Studio, the dashboard):
5
+ content and the CMS, form submissions, media, publishing and preview, SEO
6
+ and domain settings, site connections, analytics, integrations, variables
7
+ and secrets, schedules and run history, outbound webhooks, visual editing,
8
+ version history, sharing and billing, chat, and the command palette. Load
9
+ it when someone asks where something is or how to do it in Fimo, even when
10
+ they do not say "Studio", and read the matching reference before you name a
11
+ screen, button, or setting.
14
12
  ---
15
13
 
16
14
  # Fimo Studio
@@ -54,6 +52,9 @@ Read the relevant reference file when the user's question matches a topic.
54
52
  | Content Management (CMS) | references/cms.md | "How does the CMS work?", "How do I add blog posts?" |
55
53
  | Media Library | references/media-library.md | "How do I upload images?", "Where are my files?" |
56
54
  | AI Chat | references/ai-chat.md | "What can the AI do?", "How do I give context?" |
55
+ | Agents | references/ai-chat/agents.md | "Which plan has which agents?", "New agent", "Why is my agent paused?" |
56
+ | Default agents | references/ai-chat/default-agents.md | "What does the SEO agent check?", "Which routines run on Monday?" |
57
+ | Routines | references/ai-chat/routines.md | "When does a routine run?", "Routines paused", "Watch a price page" |
57
58
  | Site Knowledge | references/knowledge.md | "Set brand guidance", "Website goals", "Draft knowledge from the site" |
58
59
  | SEO Settings | references/seo-settings.md | "How do I set up SEO?", "Favicon, title, share image" |
59
60
  | Variables & Secrets | references/variables-and-secrets.md | "Where do I put an API key?", "Public or private variable?" |
@@ -72,7 +73,7 @@ Read the relevant reference file when the user's question matches a topic.
72
73
  | When user asks about... | Also mention... |
73
74
  | ----------------------- | -------------------------------------------------------------------- |
74
75
  | Publishing | SEO settings, Custom domains |
75
- | Custom domains | Publishing (must publish first), Billing (premium required) |
76
+ | Custom domains | Publishing (must publish first), Billing (Starter or higher) |
76
77
  | Adding images | Media Library, AI Chat, Edit Mode |
77
78
  | Content/CMS | AI Chat, Media Library |
78
79
  | Webhooks | Site secrets, Forms, Publishing |
@@ -0,0 +1,46 @@
1
+ # Agents
2
+
3
+ An agent is a chat with a job: a name, a one-line summary for people, and instructions that Fimo reads at the start of every reply in its chat and every run of its routines. A live agent is shared like any chat and listed under **Agents** in the chat list. What each default routine does is in `ai-chat/default-agents`; triggers, runs, reports and agent-hours are in `ai-chat/routines`.
4
+
5
+ ## What each plan gets
6
+
7
+ | | Free | Starter | Pro |
8
+ | ---------------------------------- | --------------------------------------------------- | --------------------- | --------------------- |
9
+ | Default agents | SEO and GEO | All five | All five |
10
+ | Their routines | Only **Weekly SEO checkup** and **Weekly AI check** | The detailed routines | The detailed routines |
11
+ | A change wakes a routine | Never; the checkups run weekly | At most once a day | Right after it |
12
+ | Agents of your own (**New agent**) | Set up only | Set up only | Live, with routines |
13
+
14
+ - On Free, an agent's settings list the routines Starter adds, with **See plans**.
15
+ - On Starter, a routine that ran on a change runs again on a change only a day after that run started. The changes made meanwhile wait and are checked together; the routine's next run shows when. **Run now** and schedules never wait.
16
+ - In the **+** menu, an agent the plan does not include shows **Upgrade** and opens the plans.
17
+ - Below Pro, a live agent of your own has read-only settings, says its routines are paused, and offers **See plans**. It can still be deleted.
18
+ - A plan change changes which agents open and which routines run; agents, routines and past runs stay. Fimo support can give an organization Pro's agents on any plan.
19
+
20
+ ## Create an agent
21
+
22
+ **+** then **New agent** opens the new agent's chat at once. The agent asks what it should do, with three one-click answers, then drafts its name, summary, instructions and routines on one card. Ask for changes in the chat, or select **Edit**. **Create agent** makes it live: the team sees it and its routines start.
23
+
24
+ Until then, only its creator sees it (**New agent**, **Setting up…**), its routines do not run, and its chat changes nothing on the site: it reads the site, its data and the web, with no CLI, files, work sessions or publish. It goes live only after the person saw its latest draft. When the person names no exact time for a routine, Fimo keeps the hour and picks the minute. Anyone can set up a draft; below Pro, **Create agent** opens the plans.
25
+
26
+ ## Use and change an agent
27
+
28
+ - In an agent, the header shows its name alone and the message box reads **Message <name>**. An empty default agent greets in its own words and offers starter requests; its description stays in its settings. A starter is sent as shown, by anyone who can send messages.
29
+ - Ask the agent to change its instructions; it uses its `agent` tool, and the chat shows **Updated instructions** with **View change** (before and after).
30
+ - The agent's settings have **Summary** and **Instructions** (shared with everyone; empty means none) and an **Instruction history**, where **Restore previous** puts back the text a change replaced.
31
+ - A default agent is Fimo's: its name, summary, instructions and portrait cannot change on any plan. Its routines can be switched off and on, or deleted (a deleted one stays deleted); nothing is added to it (`routine_is_default`). Routines of your own live in an agent of your own, on Pro (`routine_needs_pro`). A routine Fimo retires stops running and leaves the list.
32
+ - In text, an agent links another as `[Name](agent://id)`, shown as a chip that opens it.
33
+
34
+ ## Default agents
35
+
36
+ The **+** menu lists them under **Agents**: **SEO**, **GEO**, **Content**, **Quality**, **Compliance**. A new site starts with SEO and GEO, their routines active. Content, Quality and Compliance, and every default agent on an older site, are created the first time someone selects them, with their routines active (except Editable copy); their routines use credits. Each has one chat per site.
37
+
38
+ - **SEO:** Technical SEO, Broken links, Meta tags, Image SEO, Hreflang, Internal linking, Keyword optimization, Keyword research.
39
+ - **GEO:** AI readiness, Structured data, AI citation optimization, AI visibility.
40
+ - **Content:** Content refresh, Content lifecycle, Translation, Editable copy (off until someone switches it on), Competitors, Writing.
41
+ - **Quality:** Accessibility, Mobile optimization, Page speed.
42
+ - **Compliance:** Brand compliance, Cookie consent, Privacy and legal pages.
43
+
44
+ A change on main wakes the routines it can matter to, before a publish: CMS edits, code changes, images, languages and forms; the first publish and a new domain also wake Technical SEO and Cookie consent. A small model reads each change first, so a change that matters to no routine starts nothing; schedules run anyway. A run that a change started checks only the pages it touches.
45
+
46
+ **Alone:** a fix with one right answer from the page's own words, without a message. **Asked as a decision:** a change to what visitors read, which page wins a search, whether a page exists, a redirect with more than one possible target, a price or a legal page. They never remove a page or change the robots policy alone. A change a person asks for in chat is made. A check that could not finish (a timeout, a 5xx, a page past the crawl limit) is retried once, then reported as incomplete, never acted on.
@@ -0,0 +1,57 @@
1
+ # Default agents: what each routine does
2
+
3
+ Plans, setup and the rules every default agent follows are in `ai-chat/agents`. Each agent uses a built-in Studio skill (`fimo-seo`, `fimo-geo`, `fimo-content`, `fimo-quality`, `fimo-compliance`); the GEO, Content, Quality and Compliance site checks share `fimo-seo`'s crawl and its rule for what a run may change alone. For a site's current routines, schedules and instructions, run `fimo agents list` or `fimo agents show <name>`.
4
+
5
+ ## SEO (8 routines)
6
+
7
+ - **Technical SEO:** status, redirects, canonicals, sitemap, robots.txt, language, main heading, and what got worse since its last run: pages moved without a redirect and, with Search Console, old addresses Google still shows that the site no longer serves. Runs on changes, the first publish, a new domain, and every Monday.
8
+ - **Broken links:** links to the site's own pages that fail or pass through a redirect. Runs on changes; links to other sites on the 1st of each month.
9
+ - **Meta tags:** titles, descriptions and Open Graph previews that are missing, repeated, or too long or short (characters and pixels), and titles that do not end like the site's other titles. Runs on changes and every Monday.
10
+ - **Image SEO:** alt text, file names, and images over 200 KB. An image in the media library is rendered through Fimo's resizing; a file of the site is uploaded first. An image on another host gets a proposal for a smaller export.
11
+ - **Hreflang:** on a site in more than one language, each version lists the others.
12
+ - **Internal linking:** orphan pages, pages more than three clicks deep, and articles with no link to a hub or product page or more than eight links. Runs on changes and every Monday.
13
+ - **Keyword optimization** (every Tuesday, needs Search Console): for pages ranking 8th to 20th it proposes a title and headings, then checks 8 weeks later whether they helped.
14
+ - **Keyword research** (2nd of each month): writes at most five briefs to the site knowledge and asks which to write; the Content agent's Writing routine writes the approved ones.
15
+
16
+ The Monday runs of Technical SEO, Meta tags and Internal linking read every page of main's hosted preview, for what no single change shows (pages out of the sitemap, shared titles, title shape, orphan and deep pages). A run that a change started checks only the pages it touches, its first run too.
17
+
18
+ Alone, it fixes a link only after it checked that the new address shows the same content in the same language; it adds a link on words the text already uses to name another page, and a 301 from a page's old address once the same page is live at the new one. A title, description, heading or link it changed is not changed again for 8 weeks (Google needs 8 to 12 weeks to show the effect), unless the change broke something. The keyword routines never change the site alone.
19
+
20
+ ## GEO (4 routines)
21
+
22
+ - **AI readiness:** llms.txt, the text each page serves before JavaScript, landmarks and heading levels.
23
+ - **Structured data:** valid JSON-LD that fits each page: Organization and WebSite, Article, BreadcrumbList; Product and FAQ where they fit.
24
+ - **AI citation optimization:** a direct answer under each content page's H1, and question headings on long pages.
25
+ - **AI visibility** (once a week): how often ChatGPT, Gemini, Perplexity and Google AI Mode name and cite the site for about 20 customer questions. It changes nothing.
26
+
27
+ Alone, it writes a missing llms.txt and keeps its links in step with the pages, adds Organization, WebSite, Article and BreadcrumbList from facts the site already shows (never an invented logo, address, author or rating), and fixes landmarks and heading levels without changing a word. It asks before an opening answer, new headings, FAQ or Product data, or a change to how pages are rendered (unless one setting serves the same pages as HTML and nothing else changes). It never changes who robots.txt lets crawl, AI crawlers included, and never invents a claim; asked, it names the crawlers robots.txt keeps out.
28
+
29
+ ## Content (6 routines)
30
+
31
+ - **Content refresh** (2nd of each month): past years written as the present, events and offers whose date passed, "coming soon" text, product names the team retired, pages losing visitors in Fimo analytics.
32
+ - **Content lifecycle** (new content pages, and every three months): thin pages, pages that repeat another, and pages nothing links to that are also thin or out of date.
33
+ - **Translation:** CMS entries missing in a locale, or older than their source. On a one-language site it stops at once.
34
+ - **Editable copy:** off until someone switches it on, or runs once with **Run now**. It moves words typed in the code into labels, and lists into the CMS.
35
+ - **Competitors** (1 January, April, July, October) and **Writing** (Fridays, and when a brief is approved: it writes the page in a work session and asks for review; it never publishes).
36
+
37
+ Alone, it translates missing entries only for a type the team already keeps in that locale, and outdated entries, in the site's tone, keeping slugs, dates, numbers, references and media. A new locale or type, a price and a legal page are translated only after a person agreed. It renders a copyright year from the date when the year is typed in the code of one component or layout; a year in a CMS entry or in several places is proposed. Any other sentence a visitor reads is asked, quoting the exact words. Merging, redirecting or retiring a page is only proposed.
38
+
39
+ ## Quality (3 routines)
40
+
41
+ It also measures accessibility in each page, and opens one or two pages of each template again 390 px wide.
42
+
43
+ - **Accessibility:** WCAG AA contrast, heading levels, links, buttons and fields without a name, images used as links with no name, positive tabindex, invisible focus, the page language.
44
+ - **Mobile optimization:** a missing viewport or one that blocks zoom, sideways scrolling, images, tables and forms wider than the screen, tap targets under 24 × 24 px, text mostly under 12 px.
45
+ - **Page speed** (3rd of each month): Lighthouse on a phone for the home page and one page of each template.
46
+
47
+ Alone, it names a control or labels a field from words the page shows, fixes heading levels, tabindex and a removed focus outline, lets visitors zoom again, and keeps media inside the screen by its own width. A colour, size or layout change is asked as a decision. Image alt text stays with SEO.
48
+
49
+ ## Compliance (3 routines)
50
+
51
+ It can open a sample of pages with no consent given and record the cookies, third-party requests, consent banner and form fields a first visit gets.
52
+
53
+ - **Brand compliance:** words, terms, claims and voice of what changed, against the brand rules in the site knowledge, `AGENTS.md` and the agent's memory. It runs on Claude Haiku 5.5.
54
+ - **Cookie consent:** trackers and third-party cookies before a choice, a missing banner, a banner with no way to refuse.
55
+ - **Privacy and legal pages:** a privacy page exists, names the services loaded and the kinds of field the forms collect, and was updated in the last two years.
56
+
57
+ Alone, it changes only a term the brand rules replace. It proposes every other sentence, the consent setup and every legal page, with placeholders for legal details, and never states that a site is compliant.
@@ -0,0 +1,53 @@
1
+ # Routines
2
+
3
+ A routine is saved instructions that an agent runs in a thread of its own: by hand, on a schedule, or when something happens in the project. Each run sees the earlier runs, and the agent's chat does not fill with its work. For plans and the default agents, read `ai-chat/agents`.
4
+
5
+ ## Create and change
6
+
7
+ The agent creates, changes, runs and deletes its routines with its `routines` tool. A person asks in the agent's chat ("check the prices every Monday", "run after each publish", "pause the link sweep"). Studio only shows routines. Routines of your own need an agent of your own on Pro; a default routine can only be switched off and on, or deleted.
8
+
9
+ ## Triggers
10
+
11
+ A trigger is set in words. Fimo checks them and says when the routine runs, or why it cannot follow them and what it can do instead. Fimo can follow:
12
+
13
+ - publishes of the site (or only the first publish), and a failed publish;
14
+ - content changes on main: by content type, action, locale, changed fields, the entry's current values ("price over 100", "status changes to archived"), or a yes/no question about what the content says ("mentions Contentful or Sanity"). For "a competitor" it uses the names given or those in the site knowledge, and asks otherwise. A run started on an uncertain answer says so;
15
+ - media files uploaded or changed (by name or file type);
16
+ - the site's code changing through a deploy or a merge (by file path glob, or a yes/no question such as "a page was added");
17
+ - another routine of the same agent finishing, optionally with a yes/no question about its report (never its own end);
18
+ - a visitor submitting a form, optionally with a yes/no question about the answers; a form or content type changing; a branch merged into main; a locale added, removed or made the default; a domain added or removed;
19
+ - the prices on a page of another web site ("tell me when https://example.com/pricing changes its prices"). Fimo checks the page **once a day**; a change must hold on three reads in a row before the routine runs, and the run lists each plan with its old and new price. The routine page shows the watch as **OK**, **Needs repair**, **Blocked** or **Unstable**, with its last check. **A project can watch at most 10 pages.**
20
+
21
+ Comparing a field's old and new values ("the price goes up by 10%") is not supported. For what Fimo cannot see as an event (traffic, rankings, errors), the agent offers a schedule whose run checks it and stops when there is nothing to do.
22
+
23
+ **Chains:** what a routine changes, publishes or reports can start the agent's other routines, in chains of any length. Each routine runs at most once per change that started the chain, so routines never loop or multiply. A person's new change, or a run by hand or on a schedule, starts a new chain.
24
+
25
+ **Waiting events:** events that come in while a run is busy wait, then run together in the next run; the Runs list shows one run with its event count. A very large burst, such as a bulk import, runs in several runs, in order. The agent can make a routine's event runs wait until the changes stop for a set number of minutes (at most 2 hours from the first), then read them in one run. A yes/no question is then answered once for all the waiting changes, each entry once in its latest version; a change it says no to leaves the queue, and when it says no to all, no run starts. Default routines that read what changed wait 10 minutes after the last change (Technical SEO, Meta tags, Image SEO, Internal linking, AI readiness, Structured data, AI citation optimization, Content lifecycle, Translation, Accessibility, Mobile optimization, Brand compliance); Editable copy waits 30 minutes; those that guard the live site (Broken links, Hreflang, Cookie consent, Privacy and legal pages) run at once.
26
+
27
+ ## Where routines show
28
+
29
+ The agent's settings list its **Routines** as cards: name, when it runs (the event trigger's title, **Every Monday at 08:00**, or **By hand**), next run, and an **Active** switch. A new site's default routines wake on changes from five minutes after the site is created, so its first push and seed start no run.
30
+
31
+ A routine's page is read-only:
32
+
33
+ - **What it does**: one or two plain sentences (the agent writes them with the routine; Fimo writes a default routine's), with **Instructions** folded under it. A routine created before descriptions shows only its **Instructions**.
34
+ - **Model**: the model its runs use. Opus 5.5 is the default; most default routines run on Claude Haiku 5.5 or Claude Sonnet 5.5. A default routine's model cannot change; for a routine of your own (Pro), ask the agent to set another.
35
+ - **Triggers**: **By hand** with **Run now** (starts a run at once); **Schedule** and the event trigger when set, the trigger with a short title from Fimo and the words it was set with.
36
+ - **Runs**: newest first, two lines each. How it ended: **Quiet**, **Sent to agent**, **Asked for a decision**, **Incomplete** (it could not check what it was asked to; the agent is not told), or **Failed** and **Stopped** in red. When and what started it: **On schedule**, **On event**, or **By hand** with the person's name or **agent**. Then the first sentence of what it told the agent. Select a run to open the routine's thread, read-only, at that run.
37
+ - The **Active** switch in the header pauses the schedule and the trigger, not **Run now**.
38
+
39
+ ## Reports
40
+
41
+ Routines report to the agent, never to a person. A run with something the agent should know writes it to the agent's chat as a message only the agent reads; the agent takes a turn on it (reports that wait together are read in one turn). A run with nothing to say reports nothing.
42
+
43
+ The agent decides what the team sees: nothing (the chat and its row stay as they were), one short message (several reports folded into one), or a form when a person must decide. While that question is the last message, the chat's orange dot says **Needs you** and the list filter finds it under that status. A person answers the form in the chat; the agent sends the decision to the routine with its `routines` tool, and the routine runs again with it. Only the agent answers a routine: its page has no answer box.
44
+
45
+ Quiet runs leave no message. They fold into one line where the last run happened (**N runs since <date>**), counted from the last report the agent answered visibly. In the chat list, an agent's row shows the start of its newest message; before its first message it reads **First check <day> <time>**, **Runs when the site changes** (event triggers only) or **No checks scheduled**. A routine whose runs keep failing, or are refused for credits, is reported to the agent once per streak, as a message from Fimo.
46
+
47
+ ## Credits and agent-hours
48
+
49
+ - A routine starts from a schedule or an event only when the organization has enough credits for one usual run, and at least 300. Otherwise that run is skipped.
50
+ - Routine runs use the plan's agent-hours, from each run's start to its end; Studio chats do not. **Billing** shows the hours used this period, and each run shows its duration and credits.
51
+ - On Starter and Pro, a billing admin can set an extra usage budget: routines keep running past the included hours at $0.80 per extra hour, paid from that budget.
52
+ - When the included hours and any extra budget are used, nothing new runs: no schedule, event or **Run now** (the `routines` tool tells the agent), and a run in progress stops 15 minutes later. Studio shows **Routines paused** under the chat list and in the agent's settings, with the reset date and **Upgrade** to resume them sooner. The routine's page says until when.
53
+ - No run lasts longer than 2 hours.