Ask (General Q&A)

Updated on Oct 1, 2026

frg ask answers a question about your codebase without requiring an Issue ID, PR ID, or Ticket ID. Use it to explore architecture, conventions, and implementation details; it is read-only. To make changes, use frg run.

Prerequisites

A valid Telerik license. See Licensing.
A supported coding agent, installed and authenticated. Check it with frg status.
No work-item context is required.

Subcommands at a glance

frg ask has no subcommands. Every workflow group also has its own ask subcommand that answers with that group's context. Pick the one that matches your question:

Command	Answers questions about	Required context	Output	Default role
`frg ask`	The codebase in general	None	`answer.md` in the ask directory (see below)	`software_engineer`
`frg issue ask`	The current issue	Issue ID	`.forge/work/job_<issue_id>/issue/answer.md`	`product_manager`
`frg code ask`	The implementation and plan	Issue ID	`.forge/work/job_<issue_id>/code/answer.md`	`software_engineer`
`frg pr ask`	The pull request	Issue ID, PR ID	`.forge/work/job_<issue_id>/pr/pr_<pr_id>/answer.md`	`software_engineer`
`frg docs ask`	Project documentation	Issue ID	`.forge/work/job_<issue_id>/docs/answer.md`	`technical_writer`
`frg backlog ask`	Backlog strategy	None	`.forge/work/backlog/answer.md`	`product_manager`
`frg ticket ask`	The current support ticket	Ticket ID	`.forge/work/ticket_<ticket_id>/answer.md`	`sre`
`frg sec ask`	Security findings and remediation	Security scanner	`.forge/work/<job_id>/answer.md`	`security_analyst`

Where `frg ask` writes files

frg ask resolves the directory for question.md and answer.md from the current context:

text

Issue ID set?  ── yes ──► .forge/work/job_<issue_id>/ask/
      │ no
Ticket ID set? ── yes ──► .forge/work/ticket_<ticket_id>/ask/
      │ no
      └────────────────► .forge/work/ask/

Typical flow

text

question ──► frg ask ──► answer.md ──► follow-up frg ask (same session) ──► frg ask --clear (new topic)

Ask a broad question:

bash

frg ask "What is the overall system architecture?"

Read the answer (no Issue ID or Ticket ID set):
bash
```
cat .forge/work/ask/answer.md
```

Follow up with details, pointing the agent at relevant files:

bash

frg ask --context-file docs/architecture.md "How is authentication implemented?"

Start a fresh session when you change topic:

bash

frg ask --clear "How are database migrations run?"

frg ask

text

frg ask [OPTIONS] [QUESTION]

Option	Description	Default
`[QUESTION]`	Question to ask. If omitted, Forge reads `question.md` from the ask directory.	None
`-a, --agent <NAME>`	Coding agent for this run: `claude_code`, `github_copilot`, or `opencode`.	Agent in `agents.toml`
`-r, --role <ROLE>`	Overrides the default role: `product_manager`, `software_engineer`, `software_architect`, `technical_writer`, `security_analyst`, `sre`. Cannot be combined with `--custom-agent`.	`software_engineer`
`--custom-agent <NAME>`	Selects a custom agent configuration inside your coding agent. Cannot be combined with `--role`. Forge does not check that the name exists.	None
`-m, --model <MODEL>`	Overrides the model configured in `agents.toml`. Valid values depend on the agent.	Model in `agents.toml`
`-c, --context-file <FILE>`	Adds a file path as context. Repeatable. Relative and absolute paths work.	None
`--context-dir <DIR>`	Adds a directory as context. Repeatable. Up to 100 files per directory and 1 MB per file; binary files and hidden directories such as `.git` and `.forge` are skipped.	None
`--clear`	Clears the existing agent session and starts fresh.	Off
`--retry [<INSTRUCTIONS>]`	Continues the previous session to complete missing outputs. Requires an existing session; cannot be combined with `--clear`.	Off
`--auto-retry`	Retries once automatically when expected outputs are missing. Cannot be combined with `--clear`.	Off
`--print-prompt`	Prints the compiled prompt without running the agent.	Off
`--tail`	Follows agent execution output in real time.	Off
`--bypass-safety-checks`	Skips prompt-injection validation. Bypasses are logged. See Prompt Safety.	Off
`-q, --quiet`	Suppresses all output except errors.	Off

Examples:

bash

# Choose a role that matches the question
frg ask --role security_analyst "Are there vulnerabilities in the auth flow?"

# Several context sources
frg ask --context-dir docs/patterns --context-dir examples/ "Which patterns should I follow?"

# Inspect the prompt without running the agent
frg ask --print-prompt "Explain the caching strategy"

Ask a long question from a file

For questions with several paragraphs or code, write question.md and run frg ask without an argument. This example assumes no Issue ID or Ticket ID is set:

bash

mkdir -p .forge/work/ask
cat > .forge/work/ask/question.md << 'EOF'
# Caching strategy

We need a caching layer for the API.

1. Which caching libraries does this project already use?
2. Which existing patterns should a new cache follow?
3. How should cache invalidation be tested?
EOF

frg ask --role software_architect
cat .forge/work/ask/answer.md

When you pass QUESTION on the command line, Forge also writes it to question.md.

Troubleshooting

No question provided

Cause: You ran frg ask without QUESTION, and no question.md exists in the ask directory.

Solution:

Pass the question: frg ask "What does this project do?", or
Create question.md in the ask directory for your context (see Where frg ask writes files).

Question file exists but is empty

Cause: question.md contains only whitespace.

Solution: Add the question to the file, then run frg ask again.

Invalid context file(s): or Invalid context directory/directories:

Cause: A --context-file or --context-dir path does not exist, is not readable, or (for --context-dir) is not a directory.

Solution:

Correct the path or remove the option. Paths resolve relative to the current directory.
Run the command again.

The --retry flag requires a previous session to continue

Cause: You used --retry before any session exists.

Solution: Run frg ask without --retry first. Use --retry on a later run when outputs are missing.

The answer misses expected context

Solution:

Run frg ask --print-prompt "<question>" to see what the agent receives.
Add the missing files with --context-file or --context-dir.
Run frg status to check the agent and model.