Release v0.8.21 · FiveTechSoft/CCHarbour

CCHarbour v0.8.21 — /loop fixed-interval rerun + /rewind (double-tap Esc).

New since v0.8.20

/loop slash command. Re-run a prompt on a fixed interval,
matching Claude Code's fixed-interval /loop. /loop 5m check CI
arms a recurring turn every 5 minutes; intervals accept the s,
m, h suffixes (a bare number is seconds). /loop /
/loop status shows the active loop, /loop stop ends it while
keeping the prompt text, /loop clear drops both. The sleep
between turns is interruptible — pressing Esc ends the loop.
Hooks in parallel to the existing /goal auto-continue, so
/loop and /goal can be armed independently. Model-paced
(self-scheduled) variant is not implemented in this release.
/rewind slash command + double-tap Esc. Undo the last
conversation turn. /rewind pops one snapshot; /rewind N pops
N. A double-tap of Esc (≤ 600 ms apart) at the idle prompt is
detected in CCPROMPT_Poll and converted into a rewind
interrupt kind, which CCREPL_PromptIdle returns as the literal
/rewind. Snapshots are pushed before each user message,
/init, /btw drain message, and /loop runner turn; the
stack is capped at 20 entries (oldest falls off the bottom).
Restored state covers aMsgs, s_cGoal, s_lGoalLooping,
s_cLoopPrompt, s_nLoopIntervalSec, s_lLoopActive,
s_lPlanMode, s_lLeanMode, s_hSessionUsage, and
s_lCompactNudged. /clear and /load flush the stack. File
modifications made by write / edit / shell are NOT rolled
back — only the conversation.

v0.8.20 — Ask-prompt deadlock fix: non-TTY auto-deny + `CCHARBOUR_ASK_TIMEOUT`.

New since v0.8.19

Non-TTY auto-deny on ask prompts. When a tool gated ask
fires and stdin is not a TTY (piped input, script -c, a
background SSH exec_command, a CI runner), CCREPL_AskPerm
used to call the line reader anyway and block forever — no human
was at the keyboard to type y. Now the path is short-circuited
with a one-line [non-interactive stdin -- '<tool>' denied]
notice and the call is denied immediately, so the agent loop
surfaces an Error: tool ... denied by user result and keeps
moving instead of hanging.
CCHARBOUR_ASK_TIMEOUT env var. Bounds the wait even on a
real TTY. Set to a positive integer number of seconds; if no
answer is typed before the deadline, the REPL prints
[no response in Ns -- denied] and the call is denied. Default
(unset or 0) keeps the original blocking behaviour. Both safety
nets fall back to deny — a missing human can never silently
approve.
New CCREPL_ReadLineTimeout(nSecs) helper. Poll-based stdin
reader that wakes every 500 ms via the new
CCCON_StdInWait( nMs ) C primitive (POSIX select() on
STDIN_FILENO in src/ccconsole_posix.c; WaitForSingleObject
on the Win32 input handle in src/ccconsole.c). Same paste/echo
conventions as CCREPL_ReadLine so the editing experience is
unchanged when the user does type.

v0.8.19 — /compact slash command + auto context-fill warning.

New since v0.8.18

/compact slash command. Summarises the older part of the
conversation into one synthetic system note, keeping the system
prompt and the last 4 turns verbatim. A stateless one-shot call
with a strict "preserve exact paths / identifiers / errors /
code verbatim, no paraphrase, no preamble" instruction. Refuses
to compact when the very last assistant turn has a dangling
tool_call (no matching tool response yet), so a compaction
cannot orphan a tool-call id mid-cycle.
Soft "/compact" hint when context fills up. After every
successful turn CCREPL_MaybeWarnCompact compares the last
prompt_tokens against compact_threshold (default 0.7) of
the model's context window and prints
[context X% full -- run /compact ...] one time. Re-armable on
/clear or after a successful /compact. Never auto-runs.
Per-model context windows. CCREPL_ModelContext maps the
active model to its context size (deepseek 128k, kimi-k2 200k,
gpt-5 400k, ...) with a 32k fallback.
compact_threshold setting. New 0.7 default in
CCSETTINGS_Defaults; users can tune it per project.
dispatch_agent_background ungated. Added to the never-gated
list in CCPERM_Decide alongside dispatch_agent /
propose_agents — same consent model.

v0.8.18 — Background subagents: dispatch_agent_background tool + /tasks slash command.

New since v0.8.17

dispatch_agent_background tool. Fire-and-forget variant of
dispatch_agent. Returns a task-id (bg1, bg2, ...) IMMEDIATELY
without blocking the parent. A fresh hb_threadStart worker runs
CC_AgentRun in the background and writes progress into an in-
memory registry; the parent agent can chain its next step right
away.
Background-task registry (src/ccbg.prg). Mutex-protected
hash keyed by task id. Each record carries id, type, prompt,
timeout, status (queued / running / done / failed /
cancelled / timed_out), started_ms, ended_ms,
iterations, reply, error, cancel_requested. Public API:
CCBG_NextId / Add / Update / Get / List / Kill / CancelRequested / ClearFinished. Workers MUST NOT touch the terminal (that would
corrupt the dynamic input box); the main REPL polls the registry.
/tasks slash command. The user's only window into the
registry:
- /tasks -- tabular list (id, status, elapsed, type, prompt
  summary).
- /tasks view <id> -- full record including reply / error.
- /tasks kill <id> -- request cancel; worker exits at next
  agent-loop boundary.
- /tasks clear -- drop every finished / failed / cancelled
  / timed-out record.
Subagent filter widened. CCTOOLS_FilterForAgent removes BOTH
dispatch_agent and dispatch_agent_background from a subagent's
registry, so a subagent cannot spawn its own subagents (no
unbounded recursion).
Help, README and pages/commands.md updated with the new
command rows.

v0.8.17 — /save and /load now round-trip the full session state: goal, modes, skills, timer, and the pending suggestion.

New since v0.8.16

/save / /load preserve every REPL-level static, not just the
conversation. Saved JSON now carries a new state block plus a
suggest string in addition to model / usage / messages.
/load restores them on top of the conversation so a reload is
truly a full snapshot.
state block. CCREPL_StateExport() packs the current
goal, goal_looping, session_turn_ms, plan_mode, lean_mode,
and skills (active skill names) into a primitive hash.
CCREPL_StateImport( hState ) reapplies them, silently skipping
missing keys so legacy session files still load cleanly.
suggest field. The pending "Suggested next:" prompt that the
model emitted on its last turn is persisted alongside the messages
and re-seeded into the editor on the first idle frame after
/load, so the green Tab-acceptable preview survives a restart.
CCSKILL_ClearAll(). New helper that drops every active skill
in one shot; called by CCREPL_StateImport before re-activating the
loaded skill list so the restored set takes over from whatever was
active.

v0.8.16 — /goal turns into "keep working until the condition is met": auto-continue loop + GOAL COMPLETE sentinel.

New since v0.8.15

/goal redesigned around "keep working until the condition is met".
Setting a goal now arms an auto-continue loop in the main REPL:
after every user-initiated turn the loop scans the last assistant
reply for a literal GOAL COMPLETE sentinel. If absent it auto-
feeds Continue toward the goal. ...reply with ONLY the literal sentinel... as the next user message and runs another turn.
GOAL COMPLETE sentinel. The injected system note teaches the
model to emit GOAL COMPLETE on its own line when the condition
is met. CCREPL_GoalDone( cReply ) checks the substring.
CC_GOAL_AUTO_CAP = 25. Hard cap on auto-iterations per user
turn so a stubborn model cannot loop forever; the REPL prints
[goal auto-continue cap (25) hit ...] when it trips.
Esc pauses the loop. CCREPL_RunGoalLoop drains the
CCPROMPT_Interrupted flag and disarms s_lGoalLooping without
dropping the goal text -- [goal auto-continue paused by Esc -- /goal <text> or a new message to restart].
/goal stop. New sub-command that pauses the auto-continue
loop without clearing the goal text (so the badge stays on).
/goal clear still drops the goal entirely and disarms the loop.
Per-iteration progress line. [goal auto-continue N/25] prints
before each automatic turn so the user can see the loop ticking.

v0.8.15 — /goal slash command: pin a session-wide objective the agent carries through every turn.

New since v0.8.14

/goal slash command. Pins a session-wide objective the agent
carries through every subsequent turn until it is changed or
cleared:
- /goal — show the current goal (or (none)).
- /goal <text> — store the goal AND inject it into the
  conversation as a system note, so the model sees the intent on
  the very next request without re-sending the goal on every
  turn.
- /goal clear (alias /goal off) — drop the goal and tell the
  model to stop treating it as a constraint.
[goal] badge. When a goal is set, the status line under the
input box shows a [goal] tag alongside [plan-mode] / [lean].
/clear resets the goal too. Resetting the conversation also
clears the session goal and the session-turn timer accumulator, so
the next session starts blank.

v0.8.14 — Shell countdown lands above the box, diff bars equal width + bright-white text.

New since v0.8.13

Shell command countdown no longer painted inside the input box.
CCTool_ShowCountdown / CCTool_ClearCountdown route through the
CCREPL_OverwriteAtAnchor helper when the persistent box is
mounted, so timeout 300s · 252s left lands on the scroll-region
anchor above the box and a trailing CCREPL_Out( Chr(10) ) bakes
the final value in for the result summary. The plain-CR fallback
is kept for cooked / non-VT terminals.
Diff bars are all the same width. CCUI_ResultBlock scans the
diff once for the widest added/removed line (floor 110), then pads
every coloured bar to that width via CCUI_DiffPad( cLine, nWidth )
-- no more ragged edges when one line is longer than the rest.
Diff text rendered in bright white. Added lines go from SGR
37;42 to 97;42, removed from 48;5;52 to 97;48;5;52 -- the
red bar now matches the green bar's contrast.

v0.8.13 — Dynamic-box paint hardening, live timing in the spinner / token bar, /provider key picked up on the next turn.

New since v0.8.12

Banner anchored at row 1. CCREPL_Run emits ESC[H ESC[2J
before printing the banner so it is always at rows 1..N and the
CCPROMPT_Activate( nHeaderRows + 1 ) anchor lands exactly below it.
Previously the banner started at the cursor's row (often row 2 after
the shell cc line), so subsequent writes overwrote the banner's
bottom border.
Box paint uses absolute cursor jumps. CCPROMPT_Redraw now
positions every row of the four-row box with an absolute ESC[<row>;1H
instead of chaining CRLFs. The CRLF chain scrolled the screen up by
one row each time the box sat near the terminal bottom, which is how
pressing Esc at the idle prompt could stack 10+ leftover ╭ frames
above the box.
Wipe rule rewritten. The old-box-frame wipe in CCPROMPT_Redraw
is now [oldBoxTop .. min(oldBoxTop+3, newBoxTop-1)] minus the rows just written. CCREPL_Out stashes last_write_start and a trailing
-LF flag so Redraw knows where the chunk actually wrote. Fixes the
FlushPending bullet ("\n + glyph + 2sp", no trailing LF) being
wiped right after it was painted, and the stale top-frame left
behind by multi-line writes (paragraph breaks in the streamed reply).
Banner scrolls when the box pins. CCPROMPT_Region sets
scroll_top to 1 once box_top == nFloor, so the VT scroll region
expands up to the top of the screen and the banner finally rolls off
the top edge as new content arrives. While the box is still
travelling the banner is still pinned (current behaviour).
Spinner shows elapsed seconds. iteration_start stamps a wall-
clock in oRender[ "spinnerStartMs" ]; CCREPL_SpinnerShow appends
Ns to the spinner line (e.g. ⠦ Thinking... [32 tok] 7s).
Token bar shows turn / session seconds. CCREPL_RunTurn times
each turn with hb_MilliSeconds(), accumulates into a session-wide
static, and passes the per-turn ms to CCREPL_ShowTokenBar. The bar
now reads ▒ tokens in: 3670 out: 83 total: 3753 turn: 1.4s session: 7.2s. /clear resets the session counter.
dispatch_agent timer ticks down. The opening Agent block prints
the separator / header / prompt and then drops a live elapsed-time
line via the new CCREPL_OverwriteAtAnchor helper. The
interrupt_check callback refreshes the line at ~2 Hz so the user
sees Ns elapsed / Ms timeout · press Esc on the input box to cancel
tick in place. A final CCREPL_Out bakes the value in and lets the
"Agent done in X.Xs" line land below.
/provider key picked up on the next turn. CCCFG_Resolve now
falls back to $CCHARBOUR_CONFIG (or .ccharbour/settings.json)
when the env vars are empty, so a freshly-saved api_key reaches the
next chat call without rebuilding the client.

v0.8.12 — Start without an API key, configure the backend interactively with /provider.

New since v0.8.11

Starts even with no API key configured. The session no longer
exits with an error when DEEPSEEK_API_KEY is missing; the banner
and the input box come up normally and a yellow warning under the
banner tells the user to set up a backend.
/provider slash command — configure the backend at runtime:
- /provider — show current base_url / model / key state and
  list the four presets.
- /provider deepseek|glm|moonshot|openai — apply the preset's
  base_url and default model, persist to settings.json, and
  rebuild the API client in place.
- /provider key <secret> — store the API key in settings.json.
- /provider model <name> — switch the model only.
- /provider clear — wipe the stored API key.
Turn skipped (not crashed) when no key. Submitting a message
before a key is configured prints the warning again instead of
erroring out.
CCSETTINGS_Save is now public so /provider (and any future
runtime configuration) can persist updates without poking the file
directly.

v0.8.11 — Dynamic input box stability fixes.

New since v0.8.10

Scroll region spans the full band regardless of where the dynamic
box currently sits (scroll_bottom = nFloor - 1 always), so an early
write between banner and box does not scroll content out of the
banner.
Interactive selectors force-pin the box before painting. ask_user
and propose_agents now call a new CCPROMPT_ForcePin that drops the
box to the floor first, giving the selector a stable position right
above it and avoiding overlap with a still-travelling box.
Wipe never erases the just-written content. The old-box-frame
wipe in CCPROMPT_Redraw is now clamped to
[max(oldBoxTop, contentRow) .. newBoxTop - 1], so the rows that
hold the new user echo or the streamed reply are preserved.
Visual-row counting (auto-wrap). CCREPL_Out now advances
content_row by the visual rows a chunk consumes (LFs + wrapped
printable runs + ANSI sequences skipped), not just \n count, so a
long line that wraps doesn't sit under the redrawn box.
Clear-to-end-of-line on every write. Each line break in a
CCREPL_Out chunk is preceded by ESC[K; trailing junk on a row
(such as the right tail of the old box top frame) is wiped, so a
short content line never leaves a > hola─────────╮ artifact.

v0.8.10 — Dynamic input box, native CC logo gradient, banner adapts to terminal width, clean exit, playground polish.

New since v0.8.9

Dynamic input box position — the box no longer hard-pins at the
bottom of the terminal. After the banner paints it sits right below
the logo; each agent reply pushes it down one row per line. When the
box reaches the floor it "pins" and the scroll region takes over so
later output scrolls between the banner and the box. The banner stays
visible for as long as content allows.
Native CC logo gradient — the six-row block-drawing logo in the
banner renders with a per-character magenta → violet truecolor
gradient (fuchsia-300 → violet-600). Falls back to plain padding
when colour is off.
Banner adapts to terminal width — CCUI_Banner reads
CCREPL_Cols() and sizes its inner frame to fit, clamped to
[80, 200], with the left/right panels rebalanced so the logo stays
centred.
Clean exit cursor — CCPROMPT_Teardown now wipes the four box
rows and parks the cursor at box_top before returning. The shell
prompt that takes over appears right under the agent's last visible
output instead of below an orphan box.
Playground premium UX additions — a 💡 tip line cycled from a
pool of 10 entries (shown at startup and after Reset), a
Suggested-next pre-fill in the input (teal-italic; Tab accepts, any
printable key replaces, Enter sends), and a Close button on the
Settings panel that hides it without committing the typed values.
System prompt now instructs the agent to emit a Suggested next:
line, which the playground strips from the visible reply once it
has been lifted into the input.
Playground subagents — JS ports of dispatch_agent (reuses
runAgent with isolated messages + filtered tool registry) and
propose_agents (glass modal with checkbox per row, A/N bulk
toggle, Enter/Esc shortcuts). Browser toolset is now 14 tools.
CC logo cyan → violet in the playground (matched the native
gradient direction).

v0.8.9 — Multi-provider API support, modernised docs site, premium playground theme.

New since v0.8.8

Multi-provider API keys — CCCFG_Resolve now tries
DEEPSEEK_API_KEY → CCHARBOUR_API_KEY → GLM_API_KEY /
ZHIPU_API_KEY → MOONSHOT_API_KEY → OPENAI_API_KEY →
settings.json in order. Any OpenAI-compatible Chat Completions
endpoint (DeepSeek, GLM-4.6 / Zhipu, Moonshot Kimi, OpenAI, …) works
once base_url and model are set in settings.json.
Docs site overhaul — MkDocs Material now ships with a light/dark
palette toggle, sticky tabbed navigation, search suggest /
highlight / share, code copy + annotation, content tabs, social
icons, and a richer pymdownx set (tabbed, details, mark, keys,
tasklist, tilde, caret, betterem). The landing page becomes a hero
with badges, an ASCII screenshot, a card grid of features, a
tabbed 30-second tour, and use cases. Getting Started is rewritten
per-OS (Linux/macOS, cmd, PowerShell) with first-session walkthrough
and per-platform build instructions. Configuration adds a Providers
section with sign-up links, indicative pricing per 1M tokens, and
coding-tier ratings.
Playground premium theme —
- Animated mesh-gradient background that drifts subtly behind the
  glass panels.
- Glassmorphism on banner, footer, settings and tool cards
  (backdrop-filter blur + saturate).
- CC logo rendered with a cyan → violet gradient and an 8 s
  background-position shimmer.
- Pill-shaped version chip, pulsing "online" status dot, gradient
  border-mask on the footer, breathing focus glow on the input.
- Primary button is a teal linear gradient that lifts on hover;
  Settings and Reset get an outline "ghost" look.
- Mobile breakpoint at 760 px.
Playground gains tools — JS ports of memory (localStorage),
todo_write (with id / active_form / blocked_by) and
ask_user (in-page modal with radio options, Other free-text,
Cancel / Confirm, Esc / Enter shortcuts). The browser playground
now exposes file / web / github / memory / todo_write / ask_user
tools.
README refresh — centred hero with badges, "Why CCHarbour"
highlights, quick start, 30-second configuration guide. The
in-source feature inventory is updated from "Done (v0.5.0)" to
the current v0.8.9 set, listing all 16 builtin tools, skills,
plan / lean modes, subagents, paste collapse and dynamic box.

v0.8.8 — propose_agents gate, paste collapse, dynamic input box, bullet split, broader list of skills, and other UX polish.

New since v0.8.7

propose_agents tool — the agent batches 2+ planned subagents and
the user reviews them in an interactive multi-row selector. Space
toggles each row, A / N flip all on/off, Enter approves the filtered
list, Esc cancels the batch. The returned JSON is iterated by the
agent to call dispatch_agent once per approved item. Inherently
consented (the user is the gate).
Second-dispatch interceptor — if the agent calls dispatch_agent
twice in a row without going through propose_agents, the second call
is rejected with a redirect message asking it to batch via
propose_agents. Approved batches top up an allowance equal to the
number of approved items, so a sanctioned batch dispatches without
the gate firing again. Both counters reset at the start of each
REPL turn.
Agent block — every dispatch_agent call now renders an
absolute-positioned block (rule, " Agent working ", short
prompt summary, timeout + Esc hint). When the subagent finishes the
block closes with "Agent done in X.Xs (N iterations)" so wall-clock
and effort are visible.
Subagent timeout + Esc cancel — every dispatch_agent accepts a
timeout_s (default 120, max 600); the subagent is interrupted at
the boundary and returns [subagent timed out after Ns]. The user
can also press Esc on the input box mid-run; the subagent returns
[subagent cancelled by user].
Paste collapse — pasting multi-line text into the box now stashes
the real content in a sidecar and shows a tidy
[pasted N lines text] placeholder. Enter expands the placeholder
back transparently; Backspace clears the paste entirely.
Dynamic input box width — CCUI_InputInnerWidth and the frame
borders read the terminal column count and adapt (clamped to
[76, 200] inner). Wider terminals get a wider box without rebuilding.
Fixed sizes only kick in below 76 cols or above 200.
Bullet-run split (CCMD) — when a model emits an entire list on a
single line joined by - markers (no newlines), CCMD_SplitBulletRun
splits it into one virtual line per item before rendering. The split
applies both mid-stream (CCMD_Feed) and at end-of-turn (CCMD_Flush).
Stricter list formatting in the system prompt — both the parent
agent and subagents are told to put each list item on its own line
with - and to leave a space around **bold** so the surrounding
whitespace survives.
5 new skills shipped under .ccharbour/skills/ — brainstorming,
writing-plans, tdd, debugging, code-review (each with EN+ES
auto-trigger patterns).
/lean — toggles a minimal system prompt (no skills section, no
CC.md, no memory.md, no narration block) to save ~500-800 tokens
per turn. [lean] badge in the status line.

v0.8.7 — Subagents: dispatch_agent tool with isolated context and filtered tool registry.

New since v0.8.6

dispatch_agent tool — the agent can now spawn an isolated
subagent on a self-contained subtask. The subagent has its own
conversation, its own (filtered) tool registry, and the parent only
receives the subagent's final reply. The parent's context stays small
while exploration / multi-file searches / focused investigations run
in the subagent.
Two agent types:
- explore (default) — read-only toolset: read, glob, grep,
  github_read, memory, use_skill. Cannot modify the codebase or
  run shell commands. Ideal for "where is X used?" and survey tasks.
- general — full toolset (write, edit, shell, github_write, web_*,
  todo_write, ask_user). Subagent can perform real work in isolation.
No recursion — dispatch_agent is always filtered out of a
subagent's registry, so a subagent cannot spawn another one (yet).
Inherently consented — dispatch_agent, like use_skill,
ask_user and todo_write, bypasses the permission gate; the user
asked for the work, the agent just delegates a piece of it.
Synchronous v1 — the parent blocks until the subagent finishes.
Parallel multi-agent dispatch is a future enhancement once the thread
model is in place.

v0.8.6 — Lean mode (`/lean`) for token-saving sessions.

New since v0.8.5

/lean — toggles lean-mode. While active, CCUI_SystemPrompt
returns a minimal prompt (no skills section, no CC.md, no memory.md,
no narration block). Per-turn input drops by ~500-800 tokens (from
~2900 down to ~2100 on a vanilla "hi"). Useful late in long sessions
when context is filling up or when working against an expensive model
(Claude, GPT) where each input token matters.
/lean off — restores the full system prompt. The first system
message in the conversation is updated in place so the next turn sees
the change immediately.
[lean] badge — appears in the status line under the input box
while lean-mode is active, alongside any other badges (plan-mode,
active skills).
Auto-cleared by /clear: a session reset rebuilds the system prompt
from scratch and always uses the lean state in effect at that moment.

v0.8.5 — Richer todo_write: ids, blockers and present-continuous active labels.

New since v0.8.4

todo_write got three new optional fields:
- id — a string identifier for the task. Other tasks can refer to
  it from their blocked_by list to declare a dependency.
- active_form — the present-continuous label rendered in place of
  text while that task is in_progress (e.g. "Running tests" instead
  of "Run tests").
- blocked_by — an array of id strings naming the tasks this one
  depends on. If any of those is still pending or in_progress, the
  blocked task renders indented, dimmed, with a ↳ glyph and the
  (blocked) suffix; once every blocker is completed, it returns to its
  normal style.
Renderer upgrade — CCUI_TodoBlock now honours active_form for
in_progress items and visually differentiates blocked items so the
dependency chain is obvious at a glance.
Backwards compatible — items that omit the new fields render
exactly as before.
Tool description carries a concrete example call so the model
picks up the new shape without extra prompting.

v0.8.4 — Plan mode, expanded skill library, wider banner and input box.

New since v0.8.3

Plan mode (/plan) — a new command puts the session in "plan mode":
the write, edit, shell and github_write tools are locked at the
permission gate, the writing-plans skill is auto-activated, and the
status line shows a [plan-mode] badge. The agent can read, grep, glob,
fetch and reason freely but cannot modify the codebase until the user
approves. /plan <text> enters plan mode AND submits <text> as the
first planning prompt; /plan accept (or /plan go//plan approve)
unlocks the gate and tells the agent to proceed step by step;
/plan cancel (or /plan off) drops the plan and unlocks.
Skills library expanded — five new process-discipline skills
shipped under .ccharbour/skills/:
- brainstorming — explore intent and design before coding (clarify
  goal, list constraints, sketch alternatives, pick one with reason).
- writing-plans — turn an approved approach into an ordered,
  verifiable, multi-step plan with a done criterion per step.
- tdd — red/green/refactor with CCHarbour-specific test commands.
- debugging — reproduce, isolate, hypothesise, verify; no guessing
  patches, root cause only.
- code-review — checklist for correctness, scope creep, security,
  test coverage, readability before commit.
  Each has triggers (English + Spanish), so the agent auto-activates the
  matching skill when its description fits the request.
Wider banner and input box — the welcome banner and the persistent
input frame grew from 99 columns wide to 123, giving the model 117
text columns inside the input box (was 93). The CC logo is unchanged
but now has more breathing room.

v0.8.3 — Unified tool-call block, ask_user selector polish, and richer mid-question controls.

New since v0.8.2

Unified tool-call block — every tool call now renders as a single
block: a cyan-violet rule the full terminal width, the tool's display
label ("Bash command", "Edit", "Read", "Glob", "Web fetch", ...) on its
own line, the primary argument (command / path / pattern / url / query)
indented in bright green, and any narration the model produced on a soft
white line below.
ask_user block — the selector now paints separator, " Ask user",
blank, question, blank, options, blank, hint as one absolute-positioned
block right above the input box. Up/Down navigate with wrap-around (Down
on the last option lands on the first, Up on the first lands on the
last); each repaint lands in the same rows so the screen no longer
jitters or scrolls.
Hint tail — every question lists Esc to cancel · Tab to amend · ctrl+e to explain in dim under the options.
Mid-question keys —
- Tab (amend) drops the highlighted option into the input box
  pre-filled, so the user can edit and submit a tweaked version.
- Esc cancels the question; the tool returns "User cancelled" so the
  model drops that line of questioning.
- Ctrl+E asks the model to explain; the tool returns a sentinel the
  model interprets as "elaborate and re-ask".
Type in the box during a question — printable keys, backspace,
cursor keys, Home/End, Delete, Shift+Enter all edit the input box even
while the selector is up. Enter on a non-empty box submits the line
(cancelling the selector and queuing the message for the next turn);
/exit or /quit quits cc immediately; /btw <text> records a mid-turn
interrupt that the agent picks up at the next boundary.
Box prompt history — Up/Down in the input box now navigate the
history, matching the cooked-mode editor.
Cursor follows the box — while the selector waits for a key, the
visible cursor parks inside the input box rather than blinking above the
options.
Content preserved — when the question block first paints, the
scroll region scrolls up by exactly the block's height so any model
output above remains visible instead of being overwritten.

v0.8.2 — Project skills: process-discipline checklists the model can load on demand, with a status line and an auto-trigger.

New since v0.8.1

Project skills — .md files under .ccharbour/skills/ describe a
checklist or set of instructions the model can pull in for the turn. Each
skill has a name and a one-line description in YAML frontmatter; the model
sees the list at session start and activates one with the new use_skill
tool.
Status line — the input box gained a fourth row that lists the
currently active skills as bracketed orange tags (e.g. [superpowers]),
so it is always clear which discipline is in effect.
Auto-trigger — a skill can declare a triggers: line of
comma-separated regex patterns in its frontmatter; if the user's input
matches, the skill activates automatically, its body is injected into the
conversation as a system note, and a [skill 'X' auto-activated] line is
printed in the scroll. No regex match → no activation.
/caveman command — first-class slash command that activates the
caveman skill (ultra-compressed terse replies). Generalises to any skill
via the same CCREPL_ActivateSkill path.
Sample skills shipped — .ccharbour/skills/superpowers.md
(brainstorm → plan → execute → verify checklist) and
.ccharbour/skills/caveman.md (terse-mode rules).
System prompt awareness — the available skills are listed in the
system prompt so the model knows what is on offer without loading every
body up front.

v0.8.1 — Input box polish: white user echo and translucent-green "Suggested next".

New since v0.8.0

User prompt echoed in white — every submitted prompt is now reprinted
in bright white in the scroll above the input box, so the transcript shows
what was asked alongside the model's reply. Mid-turn queued messages are
echoed the same way when they get handled.
"Suggested next" pre-filled in the box — the model's Suggested next:
line is loaded into the input box as a translucent green suggestion. Press
Tab to accept it as the message, Backspace/Delete to clear it, or just
start typing to replace it. The cooked-mode fallback already had this; box
mode now matches.

v0.8.0 — Linux and macOS support: cc now builds and runs on all three platforms.

New since v0.7.0

Linux & macOS builds — new cc_linux.hbp / cc_mac.hbp project files,
a build_cc_linux.sh script, and build-linux / build-mac CI workflows.
Every tagged release now ships a Windows, Linux and macOS binary.
POSIX console backend — ccconsole_mac.c renamed to ccconsole_posix.c
and shared by the Linux and macOS builds (termios/select, no OS-specific
code). Added the CCCON_Size and CCCON_KeyPending functions it lacked.
Cross-platform shell tool — the shell tool ran commands only through
cmd.exe; it now uses /bin/sh on Linux and macOS. Without this every
shell command failed on those platforms.
Auto-created settings — on first run CCHarbour writes a default
.ccharbour/settings.json so the configuration is there to discover and
edit.
Build-project fix — cc_mac.hbp was missing ccprompt.prg, leaving
the macOS build broken since the always-visible prompt landed.

v0.7.0 — Shell command timeouts with a live countdown, plus reliability fixes.

New since v0.6.0

Shell timeout — the shell tool now bounds how long a command may run.
Set it with the shell_timeout setting (default 30s), an optional per-call
timeout argument, or — when shell_timeout is 0 — an automatic
per-command estimate.
Live countdown — while a shell command runs, the REPL shows the
configured timeout and the seconds still left, updated in place.
Reliability fixes — repaired the timed-shell path: a missing
fileio.ch include (F_ERROR undefined), an undefined hb_TempFile()
link error, and a broken exit-code marker. Completion detection and the
real exit code now come from hb_processValue.
UTF-8 sanitising — tool results are scrubbed of invalid Unicode
(CC_SanitizeUTF8) so they cannot break the API request JSON.
Better API errors — API failure messages now include a dump of the
HTTP response body.
Docs — documented the shell timeout; corrected web search to DuckDuckGo.

v0.6.0 — Massive refactor: all files and functions renamed from ds* to cc* prefix.

New since v0.5.1

File renaming — all source files renamed from ds*.prg to cc*.prg (e.g. dsui.prg → ccui.prg)
Function prefix change — all public and static functions renamed from DS* to CC*
(e.g. DSUI_Version() → CCUI_Version(), DSREPL_Out() → CCREPL_Out())
Build system updated — .hbp, .bat, test files and README all reflect the new names
All 340 tests pass with 0 failures

v0.5.1 — Playground fixes and updated documentation.

New since v0.5.0

Playground fixes — removed unused parameter in web.js, fixed registry test sort order
Documentation — updated module table in README with detailed descriptions for all components

v0.5.0 — DuckDuckGo web search (no API key needed), refactored REPL, Harbour ship logo, and more.

New since v0.4.0

DuckDuckGo web search — replaced Tavily with DuckDuckGo Instant Answer API, no API key required
REPL refactor — cleaner multi-turn agent loop, fixed LoadSession bug, portable paths
Harbour-style ship logo — new project branding in the startup banner
Spinner throttle — slowed down from 30 ms to 100 ms for less flicker
Default model fix — restored to deepseek-v4-flash

Previous features (v0.1.0 — v0.4.0)

Pause tool execution with Escape, web & GitHub tools, memory tool, web playground,
conversation persistence, token cost tracking, multi-line input, Ctrl+C cancel,
animated reasoning spinner, co_author setting, diff improvements, CI/CD

Usage

REM Windows
set DEEPSEEK_API_KEY=sk-...
cc.exe

# Linux / macOS
export DEEPSEEK_API_KEY=sk-...
chmod +x cc-linux && ./cc-linux

This release ships three standalone x64 binaries: cc.exe (Windows, no DLLs
required), cc-linux (Linux) and cc-macos (macOS).

v0.8.21