Changelog
New features, improvements, and fixes across releases.
v1.6.0
Improved
- cf host: Switch from static export to ISR (Incremental Static Regeneration) — new/changed docs auto-update on page refresh without rebuild
- cf host: Add full-text search via Pagefind (replaces substring-match search index)
- cf host: Use next start instead of npx serve for ISR support
- Improve planner agent — questions and assumptions surfaced first, unknowns must be resolved before planning
Fixed
- Fix /cf-plan not asking for clarification — add mandatory "Clarify Before Exploring" step and user validation gate before finalizing plan
- Fix /cf-fix proceeding without verifying problem — add "Verify the Problem Exists" and "Confirm Approach" steps
Security
- Add prompt injection defense — layered content isolation across skills, agents, and hooks
- Central security rules in cf-help (loaded at every session start)
- Security reminder in dev-rules-reminder hook (every user prompt)
- Security context preserved in compact-marker hook (survives compaction)
- /cf-research subagent template hardened with content isolation instructions
- All 5 agents updated with prompt injection awareness (code-reviewer, planner, implementer, writer, writer-deep)
v1.5.1
Fixed
- Fix plugin cache issue — bump version to force cache invalidation for users on v1.5.0
v1.5.0
New
- Add coding-friend-cli on npm (npm i -g coding-friend-cli, binary: cf)
- Add /cf-ask [question] — lightweight codebase Q&A with auto-save to docs/memory/
- Add /cf-optimize [target] — structured optimization workflow with baseline/after measurement
- Add writer agent (haiku) — lightweight doc writer for skills that generate markdown files
- Add writer-deep agent (sonnet) — fallback for content requiring deep reasoning or long context
Improved
- cf init works in non-git directories (git steps skipped gracefully)
- CLI commands: cf init, cf host, cf mcp, cf statusline, cf update
- CLI bundles lib/learn-host and lib/learn-mcp at publish time
- CLI resolves config: local → global → defaults
- Shell tab completion auto-configured on install/update
- /cf-learn and /cf-remember now delegate writing to writer agents (skill = brain, agent = hands)
- Move agents from .claude/agents/ to agents/ (correct plugin directory structure per official docs)
Removed
- Remove /cf-update skill — use cf update (CLI) instead
- Remove /cf-statusline skill — use cf statusline (CLI) instead
- Remove /cf-init skill — use cf init (CLI) instead
- Remove /cf-learn:host — replaced by cf host and cf mcp
v1.4.0
Improved
- Top-level language config — applies to all doc-generating skills
- Configurable /cf-learn: output location, categories, auto-commit, README index
- /cf-learn supports readmeIndex: "per-category"
- Layered config: global (~/.coding-friend/) + local (.coding-friend/), local wins
- /cf-init re-runnable, asks about language/learn settings, configures permissions
- /cf-learn auto-invokes on substantial new knowledge
- Config schema in docs/config-schema.md
v1.3.0
New
- Add /cf-init — initialize workspace (docs folders + .gitignore)
v1.2.2
New
- Add /cf-update — update plugin + refresh statusline
Fixed
- Fix missing skill references in hooks
v1.2.1
Improved
- Show plugin version in statusline
v1.2.0
New
- Add /cf-research [topic] — web search + parallel subagents → docs/research/
v1.1.0
New
- Add /cf-statusline — auto-setup statusline
v1.0.3
Improved
- Change statusline from hook to statusLine setting
- Update statusline.sh for printf output
v1.0.2
Fixed
- Fix statusline JSON response format
v1.0.1
Improved
- Show active model in statusline
v1.0.0
Improved
- Initial release — 12 skills, 7 hooks, 3 agents
- Config via .coding-friend/config.json and .coding-friend/ignore