Check if there are any errors (not just warnings)

Check if the workspace contains files with any of the given extensions. Scans up to a reasonable depth to avoid slow startup on large repos.

Check if requirements file exists

Language detection functions

Highlight code with syntect.

Initialize logging based on CLI verbosity settings.

Inject completed research results into the conversation context. Returns the number of research results injected.

Install cargo-llvm-cov via cargo install

Install llvm-tools-preview via rustup

Convert to final ToolCall if valid (has a name).

Check if cargo-llvm-cov is installed

Check if an error is a recoverable connection error

Check if this message is a dehydrated stub

Check if a message is a final_output tool call

Check if a JSON parse error indicates incomplete data (vs. malformed JSON).

Check if rustup component llvm-tools-preview is installed

Check if the current plan is in a terminal state (all items done or blocked). Returns true if: - A plan exists AND all items are in terminal state (d

Check if a character could start a markdown delimiter

Check if message at index i is a result of a TODO tool call.

List all available language prompts.

List project directories from ~/projects/, sorted alphabetically.

List all available sessions in the current working directory. Returns sessions sorted by creation time (most recent first).

Load an agent prompt, checking workspace first, then falling back to embedded. Returns the prompt content and a boolean indicating if it was loaded f

Load embedded skills into the map.

Log an error to the session JSON file. Appends an error entry to the conversation history in the session log.

Log detailed error information when stream produces no content

Create a tool calls chunk (not finished).

Create a hint chunk indicating a tool call is still actively streaming. This is used to trigger UI updates (like blinking indicators) during long tool

Create a hint chunk indicating a tool call is being streamed.

Mark session as complete

Create a provider for multi-turn conversation Each call returns the next response in sequence

Parse all JSON tool calls from the accumulated buffer (used at stream end).

Parse diff stats from str_replace result

Parse a line range string like "42" or "42-118". Returns (start_line, optional_end_line).

Pass through a character without filtering (used inside code fences)

Verify a completed plan. Called when all items are done/blocked. Returns a PlanVerification with results for each done item's evidence. Blocked items

Print a compact write_envelope output showing pipeline stages. Displays: envelope written → rulespec compiled → verification → token stamped fact_grou

Print a g3 status message in progress (no status yet) Format: "g3: <message> ..." Uses centralized G3Status formatting.

Print image using iTerm2 imgcat protocol with info line.

Print a compact single-line tool output (for file operations) Format: " ● tool_name | path [range] | summary | tokens ◉ time" Returns true if the tool

Update fence state based on new streaming content.

Process a single character in AwaitingJson state

Process a single character.

Process a character for code fence detection (streaming-safe) Returns the string to emit (empty if content should be suppressed)

Process character while in a blockquote.

Process character while in a code block.

Process character while in a table.

Process italic text that may contain nested bold Matches *text* where the * is not part of

Process a single character in InToolJson state

Process a single character in MaybePattern state

Process character in normal (non-block) mode.

Process a single character in Prose state

Process streaming response from Gemini

Process a single character in InToolName state

Move new_requirements.md to current_requirements.md

Prompt user to approve refined requirements

Prompt user for codepath if not provided

Prompt user to confirm implementation completion

Prompt for new requirements