CCAR-F · Examen de práctica
60 preguntas · 120 minutos · corte 720/1000
Material no oficial. Preguntas originales escritas contra los objetivos publicados en la guía del examen. No reproducen contenido del examen real. Ver aviso completo.
Las respuestas están en respuestas.md. Para practicar con corrección automática y cronómetro, usá el simulador.
1. Selecciona 2
D2 · Claude Code: configuración y workflows
A platform team is enabling AI-assisted development across twelve squads. Which TWO measures scale the practice without scaling the risk?
- A. Keep shared configuration and prompt standards under version control
- B. Define which commands automated tooling may run and on which branches
- C. Grant the tooling broad credentials so squads are never blocked
- D. Let each squad publish its own standards independently
- E. Adopt every new assistant tool in the week it becomes available
2. Selecciona UNA
D2 · Claude Code: configuración y workflows
Reviews of AI-authored pull requests almost never raise findings, yet defects keep reaching production. What is the most likely cause?
- A. The reviewer shares the context and assumptions that produced the code
- B. The review prompt is too short to describe what quality means here
- C. The model used for the review is smaller than the one used for authoring
- D. The team merges before the automated review has finished running
3. Selecciona UNA
D5 · Manejo de contexto y fiabilidad
An agent in production starts calling an unexpected tool on a minority of requests. What does the on-call engineer do first?
- A. Read traces of the affected sessions to locate where the choice diverges
- B. Rewrite the tool descriptions, since ambiguity is the most usual explanation
- C. Disable the tool until someone can reproduce the behaviour in staging
- D. Roll the whole service back to the release from the previous quarter
4. Clasifica cada fila
D2 · Claude Code: configuración y workflows
For each situation, identify the practice that applies.
| # | Escenario | Tu respuesta |
|---|---|---|
| 1 | Two squads produce different code style from the same assistant. | |
| 2 | AI-written changes pass review, yet defects keep reaching production. | |
| 3 | An agent begins calling an unexpected tool on a subset of requests. | |
| 4 | An unattended pipeline job could run a destructive command. | |
| 5 | A new hire should inherit the team's prompt standards on day one. |
Opciones: shared configuration in version control · review from an independent context · inspect the traces first · restrict what automation may execute
5. Selecciona 2
D1 · Arquitectura agéntica y orquestación
A fixed workflow for supplier onboarding now fails on roughly a fifth of cases, where staff step in manually to finish the job. Which TWO signals would justify moving this stage to an autonomous agent?
- A. The number of onboarding requests per week has roughly doubled since launch
- B. The manual interventions follow no repeatable pattern the team can list
- C. The steps differ per supplier and only become clear part-way through
- D. Leadership wants the solution to look more advanced to the board
- E. The cost per request has risen along with the added volume
6. Selecciona 2
D1 · Arquitectura agéntica y orquestación
A supervisor coordinates four specialist agents in a regulated claims process. Which TWO responsibilities belong to the supervisor rather than to the specialists?
- A. Deciding the order the specialists run in and when to stop the chain
- B. Applying the rules that determine whether a claim is fraudulent
- C. Recording the ordered outcome of every step for the audit trail
- D. Formatting each specialist's own output for its internal use
- E. Choosing which model version each specialist should call at runtime
7. Selecciona 2
D1 · Arquitectura agéntica y orquestación
An architect is splitting a long analysis task into sequenced subtasks. Which TWO practices make the decomposition real rather than cosmetic?
- A. Passing structured output between steps so each one can be checked
- B. Giving every step only the context that step actually needs
- C. Running each of the steps on the largest model in the family
- D. Keeping the original prompt intact and appending the steps
- E. Merging the steps back into one call once quality has improved
8. Selecciona 2
D1 · Arquitectura agéntica y orquestación
Which TWO conditions indicate that a single augmented model call is enough, with no workflow or agent around it?
- A. The task is one transformation whose context can be supplied up front
- B. Several specialists must contribute before the output is complete
- C. Any reference material needed fits alongside the request itself
- D. The next step depends entirely on what the previous step happened to return
- E. Each step has to be recorded separately for an external auditor
9. Clasifica cada fila
D1 · Arquitectura agéntica y orquestación
For each situation, identify who should decide the next step.
| # | Escenario | Tu respuesta |
|---|---|---|
| 1 | The five stages of a monthly regulatory report that has not changed in two years. | |
| 2 | Which log query to run next while diagnosing a production outage. | |
| 3 | Whether to release a payment that exceeds the agreed approval threshold. | |
| 4 | The order in which four specialist agents run inside a regulated pipeline. | |
| 5 | How many sources to consult before answering an open research question. |
Opciones: the architect, at design time · the model, at run time · a human reviewer, per case
10. Selecciona UNA
D1 · Arquitectura agéntica y orquestación
An assistant handles two kinds of request: simple status lookups that arrive constantly, and complex disputes that need deep analysis. Which design serves both?
- A. A router that classifies the request and sends it down the fitting path
- B. One agent with every tool available, deciding what to do per request
- C. A single prompt long enough to cover lookups and disputes together
- D. Two separate products, each with its own interface and its own login
11. Selecciona UNA
D1 · Arquitectura agéntica y orquestación
A pipeline extracts data, transforms it and writes it to a system of record. Where does validation belong?
- A. After each step, so a bad result never reaches the following one
- B. At the very end, right before the write to the system of record
- C. Only at the start, validating the input document that arrived
- D. Nowhere: the model is instructed to check its own work as it goes
12. Selecciona UNA
D1 · Arquitectura agéntica y orquestación
In a four-stage chain, the third stage fails on about one request in twenty. What belongs in the design?
- A. A defined behaviour for that failure: retry, fall back, or stop and report
- B. A fifth stage that reviews the output and repairs whatever looks wrong
- C. A larger model in stage three so the failures become less frequent
- D. A rule telling the model to skip stage three when it cannot complete
13. Selecciona UNA
D3 · Prompt engineering y salida estructurada
A summarisation feature runs on the largest model in the family. Evaluation shows a mid-tier model matches its quality on this task at a third of the cost. What should the architect do?
- A. Move the feature to the mid-tier model and keep the evaluation running
- B. Stay on the largest model, since summary quality varies a lot with the input
- C. Split the traffic evenly between both models to hedge against a drop
- D. Move to the smallest model, since the mid-tier one is not the cheapest
14. Clasifica cada fila
D3 · Prompt engineering y salida estructurada
For each symptom, identify the technique that fixes it.
| # | Escenario | Tu respuesta |
|---|---|---|
| 1 | Output format is close but inconsistent from one run to the next. | |
| 2 | A critical rule buried in 12,000 tokens of context is followed erratically. | |
| 3 | The same 9,000-token policy block is re-sent and re-charged on every call. | |
| 4 | Answers about the middle of a large document are unreliable and expensive. | |
| 5 | The house table layout is described in prose but never reproduced exactly. |
Opciones: few-shot examples · move the rule to an edge · prompt caching · retrieve only what is relevant
15. Selecciona UNA
D3 · Prompt engineering y salida estructurada
A classification task has eight well-defined categories and clear written definitions for each. Accuracy is already high without examples. What follows?
- A. Keep it zero-shot and spend the tokens on the definitions instead
- B. Add few-shot examples anyway, since they always improve accuracy
- C. Add one example per category to make the prompt more consistent
- D. Move to a larger model to protect the accuracy already achieved
16. Selecciona UNA
D5 · Manejo de contexto y fiabilidad
A support conversation runs long and the earliest messages start falling out of the window. What preserves quality?
- A. Summarise the older turns and keep the decisive facts verbatim
- B. Let the oldest turns drop, since recent context matters more
- C. Move to the model with the largest window and keep everything
- D. Start a new conversation whenever the window gets close to full
17. Selecciona UNA
D4 · Diseño de herramientas e integración MCP
A team wants to expose forty internal endpoints to an agent. Tool-selection accuracy in testing drops sharply past roughly a dozen tools. What is the sound design response?
- A. Group the endpoints into task-oriented tools and load them on demand
- B. Register all forty tools and rely on a more capable model to choose
- C. Register all forty and instruct the model to think before selecting one
- D. Expose one generic tool that accepts the endpoint name as a parameter
18. Selecciona UNA
D4 · Diseño de herramientas e integración MCP
A retrieval index holds documents from many client organisations. What keeps one client from ever seeing another's content?
- A. Filter by tenant at query time using metadata on every chunk
- B. Instruct the model to answer only from the requesting client's files
- C. Keep one index and rely on relevance to surface the right documents
- D. Ask the model to check the client name before quoting a passage
19. Selecciona UNA
D5 · Manejo de contexto y fiabilidad
A knowledge base gets edits all day long, and the assistant must reflect them within minutes. What indexing approach fits?
- A. Incremental indexing triggered by each change, with a periodic full rebuild
- B. A nightly full rebuild, which is simpler to operate and to reason about
- C. Cache answers for an hour so the load on the index stays predictable
- D. Retrieve straight from the source database and skip the index entirely
20. Selecciona UNA
D4 · Diseño de herramientas e integración MCP
An internal API exposes forty endpoints. The team is deciding how to expose it to the agent. What should drive the tool boundaries?
- A. The tasks the agent needs to accomplish, not the endpoints that exist
- B. One tool per endpoint, so the whole surface stays available to it
- C. A single tool that receives the endpoint name and its parameters
- D. The endpoints the backend team considers most stable over time
21. Selecciona UNA
D5 · Manejo de contexto y fiabilidad
An agent platform needs alerting. Which signal deserves a page to the on-call engineer?
- A. Tool error rate crossing a threshold sustained over several minutes
- B. Any single request that took longer than the median to complete
- C. Daily token spend rising above the same day of the previous week
- D. A user rating one individual answer as unhelpful in the interface
22. Selecciona UNA
D1 · Arquitectura agéntica y orquestación
A report needs four independent sections researched and then written as one document. How should the work be split across agents?
- A. Four subagents research in parallel, one agent writes the document
- B. Four subagents each research and write their own section
- C. A single agent researches and then writes everything in one long pass
- D. Four subagents write, and a fifth merges their four drafts
23. Selecciona UNA
D1 · Arquitectura agéntica y orquestación
An orchestrator dispatches subagents whose number is only known once the problem is read. Which pattern is this?
- A. Orchestrator–workers
- B. Routing across a fixed menu
- C. Parallelization over fixed sections
- D. Prompt chaining in a fixed sequence
24. Selecciona 2
D1 · Arquitectura agéntica y orquestación
A subagent receives its assignment from an orchestrator. Which TWO parts of that assignment should a person define in advance, not the orchestrator at runtime?
- A. The output format every subagent must return
- B. The hard limits and the tools it may use
- C. The specific question this subagent must answer
- D. The portion of the problem it has been given
- E. The order in which it should consult its sources
25. Selecciona 2
D1 · Arquitectura agéntica y orquestación
A team is deciding whether a multi-agent design is worth its overhead. Which TWO conditions make it pay off?
- A. The subtasks are independent and do not need to talk to each other
- B. Each subtask produces bulk material that is better summarised apart
- C. Several subtasks must run in a strict order, each using the last result
- D. Two agents need to read and update the same shared state as they go
- E. The team wants the architecture to look more capable to the client
26. Selecciona 2
D1 · Arquitectura agéntica y orquestación
Results from parallel subagents come back in incompatible shapes and cannot be combined. Which TWO fixes address the cause?
- A. Define one output contract that every subagent must fill in
- B. Validate each subagent's return against that contract before use
- C. Ask the orchestrator to reformat whatever each subagent sent
- D. Increase each subagent's context so it can see all the others' work
- E. Run the subagents sequentially so later ones copy the first
27. Selecciona 2
D2 · Claude Code: configuración y workflows
A PreToolUse hook exits with code 2 after a blocked command. Which TWO things happen next?
- A. The tool call never executes
- B. What the hook wrote to stderr goes back to the model
- C. The session ends and has to be restarted by the user
- D. PostToolUse still runs so the block can be recorded
- E. The model retries the same command automatically
28. Selecciona 2
D2 · Claude Code: configuración y workflows
A monorepo holds twelve applications, each with its own conventions. Which TWO practices keep context usage sane?
- A. Put app-specific conventions in a CLAUDE.md inside each app folder
- B. Keep only shared, repo-wide conventions in the root CLAUDE.md
- C. Concatenate every app's conventions into the single root file for clarity
- D. Move all conventions into settings.json so they load on demand
- E. Ask developers to paste the relevant conventions at session start
29. Selecciona UNA
D3 · Prompt engineering y salida estructurada
A downstream service parses the model's output as JSON and crashes when a field is missing. What is the correct fix?
- A. Enforce the shape with a schema instead of asking for it in the prompt
- B. Add a sentence to the prompt insisting the JSON must be complete
- C. Wrap the parser in a try block and skip records that fail
- D. Post-process the text with a regular expression to patch the gaps
30. Selecciona UNA
D3 · Prompt engineering y salida estructurada
A prompt mixes instructions, a reference document and a user question, and the model keeps confusing them. What is the most direct improvement?
- A. Delimit each part with explicit tags so the boundaries are unambiguous
- B. Repeat the instructions again after the reference document ends
- C. Shorten the reference document until the confusion disappears
- D. Raise the temperature so that the model interprets the layout more loosely
31. Selecciona UNA
D3 · Prompt engineering y salida estructurada
An output will be read by a person, not parsed by a program. What follows for its format?
- A. Prose is fine; a strict schema adds cost without buying anything
- B. A schema should still be enforced for consistency across responses
- C. The output should be JSON so it can be reused programmatically later
- D. The format should change per request to suit the reader's preference
32. Selecciona UNA
D3 · Prompt engineering y salida estructurada
Few-shot examples were added to a classification prompt and accuracy on one rare category got worse. What is the likely cause?
- A. The examples over-represent the common categories and skew the model
- B. The examples are too short for the model to learn the whole pattern from
- C. Few-shot prompting does not work for classification tasks at all
- D. The temperature is too low for the examples to have any influence
33. Selecciona UNA
D3 · Prompt engineering y salida estructurada
An extraction step must return a consistent output shape. Prose instructions describing the format keep producing variations. What works better?
- A. Two to four worked input and output examples in the prompt
- B. A longer written description of the format
- C. A reminder at the end of the prompt to respect the format
- D. A lower temperature so the output varies less between runs
34. Selecciona UNA
D3 · Prompt engineering y salida estructurada
A review prompt says 'be conservative and report only high-confidence findings'. Results are inconsistent between runs. What is the defect?
- A. The criterion is vague, so each run interprets it differently
- B. The prompt is too short for the model to apply judgement
- C. The model needs a larger context to be consistent about it
- D. Confidence cannot really be expressed in a prompt in any useful way
35. Selecciona UNA
D4 · Diseño de herramientas e integración MCP
A tool wraps an internal API with eleven optional parameters. The agent fills them inconsistently. What is the better design?
- A. Expose a task-oriented tool that takes only what the task needs
- B. Document all eleven parameters more thoroughly in the description
- C. Add validation that rejects calls with unusual parameter combinations
- D. Let the agent send free-form text and parse the parameters out of it
36. Selecciona UNA
D4 · Diseño de herramientas e integración MCP
A tool returns an empty list when the underlying service times out. What is wrong with that?
- A. The model reads it as 'no results' and answers confidently from nothing
- B. The empty list uses more tokens than an explicit error would use
- C. Timeouts should always be retried inside the tool without informing anyone
- D. Empty lists are valid only when the query itself was malformed
37. Selecciona UNA
D4 · Diseño de herramientas e integración MCP
Two agents from different companies must exchange work without either exposing internal tools. Which mechanism fits?
- A. An agent-to-agent protocol, each side mediating its own systems
- B. An MCP server published by one of the companies for the other to use
- C. A shared tool catalogue registered in both agents at once
- D. A direct API integration between the two internal systems
38. Selecciona UNA
D4 · Diseño de herramientas e integración MCP
Two tools share the description 'Retrieves information' and the agent keeps picking the wrong one. What fixes it?
- A. Rename each tool and state the concrete data each one returns
- B. Add an instruction telling the model to think before choosing
- C. Merge both tools into one that takes a mode parameter
- D. Remove one of the two tools from the agent's catalogue
39. Selecciona UNA
D4 · Diseño de herramientas e integración MCP
A project's MCP server needs an API token, and the configuration file is committed to the repository. How is that handled?
- A. Reference an environment variable from the configuration file
- B. Commit the token and rotate it whenever someone leaves
- C. Keep the whole configuration out of version control instead
- D. Store the token in the project CLAUDE.md so it is documented
40. Selecciona UNA
D4 · Diseño de herramientas e integración MCP
A developer wants to try an MCP server on their own, without affecting teammates. Where does that configuration go?
- A. In the user-level configuration in their home directory
- B. In the project configuration, disabled by default
- C. In a branch of the repository that is never merged
- D. In the project configuration, with a comment explaining it
41. Selecciona UNA
D4 · Diseño de herramientas e integración MCP
An agent burns several tool calls each session just discovering what data is available. What reduces that?
- A. Expose a resource that maps the available data up front
- B. Increase the agent's context so it can remember prior sessions
- C. Add a tool that lists every other tool and what it does
- D. Instruct the agent to guess the structure before querying
42. Selecciona UNA
D5 · Manejo de contexto y fiabilidad
A long agent session degrades as it goes: earlier facts get dropped and answers drift. What is the sound response?
- A. Re-inject the facts that must survive, instead of trusting the history
- B. Increase the temperature so the model explores the history more
- C. Ask the model to summarise the conversation and continue
- D. Start over from scratch each time the drift becomes noticeable
43. Selecciona UNA
D5 · Manejo de contexto y fiabilidad
A retrieval step returns nothing relevant for a user question. What should the system do?
- A. Say it has no grounded answer and offer the next step
- B. Answer from the model's own knowledge without citing anything
- C. Return the closest passages anyway and let the user judge
- D. Retry the same query until something relevant comes back
44. Selecciona UNA
D5 · Manejo de contexto y fiabilidad
In a long support conversation, the order number and the amount keep getting lost after the history is summarised. What fixes it?
- A. Keep a case-facts block outside the history and resend it every turn
- B. Summarise less aggressively so more of the history survives
- C. Ask the model to repeat the key facts at the end of each reply
- D. Raise the context window so summarisation happens later
45. Selecciona UNA
D5 · Manejo de contexto y fiabilidad
A long input is processed reliably at the beginning and the end, but details in the middle are missed. What is the mitigation?
- A. Place the decisive information near the start or the end
- B. Repeat the whole input twice so the middle appears elsewhere
- C. Increase the temperature so more of the input is considered
- D. Split the input into random chunks and process them apart
46. Selecciona UNA
D5 · Manejo de contexto y fiabilidad
A long investigation will span several sessions. How are the findings preserved?
- A. Write key findings to a file and reference it in the next session
- B. Keep the session open indefinitely so nothing is ever lost
- C. Rely on compaction to retain the important parts of the history
- D. Repeat the discovery steps at the start of each new session
47. Selecciona 2
D1 · Arquitectura agéntica y orquestación
A coordinator spawns subagents with the Task tool. Which TWO statements about their context are correct?
- A. A subagent does not inherit the coordinator's conversation history
- B. Everything the subagent needs must be stated in its own prompt
- C. Subagents share a memory space that persists across separate Task calls
- D. A subagent can read the coordinator's context window on demand
- E. Subagents inherit the coordinator's allowed tools automatically
48. Selecciona 2
D1 · Arquitectura agéntica y orquestación
Which TWO situations call for starting a fresh session rather than continuing the current one?
- A. The tool results in context are stale and no longer reflect reality
- B. A long stretch of unrelated work has degraded the working context
- C. The next task needs the same repository and the same conventions
- D. A teammate wants to explore an alternative branch of that same task later
- E. The conversation is long but every earlier fact is still accurate
49. Clasifica cada fila
D1 · Arquitectura agéntica y orquestación
For each stop_reason value, identify what the loop should do next.
| # | Escenario | Tu respuesta |
|---|---|---|
| 1 | tool_use | |
| 2 | end_turn | |
| 3 | max_tokens | |
| 4 | stop_sequence | |
| 5 | tool_use returned after appending previous results |
Opciones: execute the requested tools and continue · show the result to the user · raise the output budget and retry · treat the custom stop sequence as the boundary
50. Selecciona 2
D2 · Claude Code: configuración y workflows
Which TWO statements about hook security are accurate?
- A. A hook runs locally with the user's own permissions, without asking
- B. A hook from an untrusted repository is arbitrary code execution
- C. Hooks run inside a sandbox that isolates them from the whole file system
- D. Hooks require explicit confirmation from the user on every trigger
- E. Hooks can only execute commands that appear in the allow list
51. Selecciona 2
D2 · Claude Code: configuración y workflows
Which TWO facts about permissions in settings.json are correct?
- A. A deny rule beats an allow rule from any other configuration layer
- B. Rules are written per tool and can use patterns to narrow scope
- C. The model reads the permission list and explains its choices from it
- D. Allow lists are advisory and the model may override them if needed
- E. Permissions are inherited only when the project has no local file
52. Clasifica cada fila
D2 · Claude Code: configuración y workflows
For each need, identify the Claude Code mechanism that fits.
| # | Escenario | Tu respuesta |
|---|---|---|
| 1 | The team's preferred naming conventions, which the model should follow. | |
| 2 | A guarantee that no destructive migration command ever executes. | |
| 3 | Declaring which Bash commands are allowed without a prompt. | |
| 4 | A specialised procedure the model should pull in when it applies. | |
| 5 | Blocking a tool call and telling the model why it was refused. |
Opciones: CLAUDE.md · hook · settings.json · skill or subagent
53. Selecciona 2
D2 · Claude Code: configuración y workflows
A CLAUDE.md pulls in other files with the @path import syntax. Which TWO statements are correct?
- A. Relative paths resolve against the file doing the importing
- B. Imports can nest, but only up to a maximum depth of five
- C. Imported files load only when the model explicitly asks for them
- D. Imports accept a space between the @ and the path for readability
- E. Only Markdown files can be imported, never configuration files
54. Selecciona 2
D2 · Claude Code: configuración y workflows
A team moves its conventions from one long CLAUDE.md into .claude/rules/ files with paths frontmatter. Which TWO gains does that produce?
- A. Each rule loads only when files matching its patterns are touched
- B. Context and tokens are spent only on the conventions in play
- C. Rules become enforceable, so the model can no longer ignore any of them
- D. All conventions load at session start, which speeds up responses
- E. The rules apply to every repository the developer works in
55. Selecciona 2
D2 · Claude Code: configuración y workflows
Which TWO statements about the built-in /compact and /memory commands are correct?
- A. Compacting can lose numeric values and dates in the summary
- B. What /memory saves reloads automatically in later sessions
- C. Compacting preserves the full transcript and only hides it
- D. Memory entries apply only to the session where they were written
- E. Compacting is required before any session can exceed its window
56. Clasifica cada fila
D2 · Claude Code: configuración y workflows
For each need, identify where the definition belongs.
| # | Escenario | Tu respuesta |
|---|---|---|
| 1 | A review checklist the whole team should share. | |
| 2 | A personal shortcut one developer uses across every project. | |
| 3 | A guarantee that a destructive command never runs. | |
| 4 | A noisy investigation routine that should not fill the session. | |
| 5 | MCP servers every contributor to this repository needs. |
Opciones: project-level, committed to the repository · user-level, in the home directory · a hook registered on a tool event · a skill with context: fork
57. Selecciona UNA
D3 · Prompt engineering y salida estructurada
An extraction schema marks every field as required. Documents that lack a field come back with invented values. What is the fix?
- A. Make fields that may be absent optional or explicitly nullable
- B. Add an instruction telling the model never to invent values
- C. Reject any extraction where a required field looks implausible
- D. Split the schema so each field is extracted in its own call
58. Selecciona UNA
D3 · Prompt engineering y salida estructurada
A classification enum keeps forcing borderline documents into the closest available category. What design change helps?
- A. Add an unclear value, plus an other option with a detail field
- B. Add more categories until every observed document fits one
- C. Instruct the model to pick the category it considers most likely
- D. Raise the temperature so the classification is less rigid
59. Selecciona UNA
D3 · Prompt engineering y salida estructurada
Invoice totals occasionally disagree with the sum of their line items, and nothing downstream notices. Which extraction design surfaces it?
- A. Extract the stated total and the calculated total, and flag conflicts
- B. Extract only the calculated total, since it is the trustworthy one
- C. Extract only the stated total, since it is what the document says
- D. Round both values so small differences stop being reported
60. Clasifica cada fila
D4 · Diseño de herramientas e integración MCP
For each need, identify the built-in tool that fits.
| # | Escenario | Tu respuesta |
|---|---|---|
| 1 | Find every test file matching a naming pattern across the repository. | |
| 2 | Locate which files mention a specific error message. | |
| 3 | Load a file in full to trace how a function flows. | |
| 4 | Change one unique line inside an existing file. | |
| 5 | List all TypeScript files under a components directory. |
Opciones: Glob · Grep · Read · Edit