v1.3.5

[1.3.5] - 2026-03-12

Bug Fix

Agent and skill files were showing as conflicts on every update

When running npx ocs-stats update, all 4 agent files and all 4 skill files always appeared under Conflicts even if the user had never modified them. The smart merge was silently skipping them and the latest instructions (including the level-up fix from v1.3.3) were never applied.

Root cause — two bugs in src/merge.js:

1. FILE_CATEGORY.SKILL was missing from SMART_MERGE_CATEGORIES — so skill files always fell through to the conflict handler instead of being smart merged
2. Path matching for agents/ and skills/ used includes('/agents/') and includes('/skills/') — the leading / never matched relative paths like agents/security.md, so agent files were also mis-categorised and fell through to conflicts

What was fixed:

• Added FILE_CATEGORY.SKILL to SMART_MERGE_CATEGORIES
• Fixed path matching to also handle paths starting with agents/ or skills/ (no leading slash)
• Added a dedicated SKILL handler in the smart merge block — skill files now update their body from the template while preserving any user frontmatter fields, identical to how agent files are handled

Result: Running npx ocs-stats update now correctly shows agent and skill files under Smart Merged instead of Conflicts, and users automatically receive the latest instructions on every update.

v1.3.4

Fix: OCS version

v1.3.3

[1.3.3] - 2026-03-12

Bug Fix

Agents on fresh installs did not know to reset XP on level-up

When installing to a new project, agents only receive the .opencode folder — not the CHANGELOG. This meant the level-up and XP carry-over behaviour documented in v1.3.0 was never communicated to the agent, so XP would just keep accumulating past level thresholds with no reset.

What was broken:

• All 4 agent files (security, testing, docs, code-review) showed a "Level Thresholds" table with cumulative XP values (0, 150, 450, 900, 1500, 3000) — inconsistent with xp.json which stores per-level XP (150, 300, 450, 1500, 3000)
• No instruction anywhere told the agent to reset xp to 0 on level-up or carry overflow forward
• All 4 skill files (SKILL.md) had the same gap — "Update XP" step with no level-up logic

What was fixed:

• Corrected the Level Thresholds table in all 4 agent files — column renamed from XP Required (cumulative) to XP to Next Level (per-level), values now match xp.json
• Added a mandatory Level-Up Rule block to all 4 agent files and all 4 skill files with explicit step-by-step instructions:
  1. Subtract xpToNextLevel from total XP to get carry-over
  2. Increment level, update title
  3. Set xp to carry-over (never accumulate across levels)
  4. Save to xp.json
  5. Chain if carry-over also exceeds the next threshold

Example now documented directly in every agent and skill file:

Level 1 (xpToNextLevel = 150), earn 230 XP → carry-over = 80 → Save: level = 2, title = "Apprentice", xp = 80

v1.3.2

[1.3.2] - 2026-03-10

Bug Fix

Stats display was showing zeros for everything

When you ran npx ocs-stats security, npx ocs-stats testing, npx ocs-stats code-review, or npx ocs-stats docs, all stats showed 0 and the progress bar always showed MAX LEVEL! — even when you had real progress.

This is now fixed. Stats correctly show your actual progress, and the XP bar shows the correct progress toward your next level.

What was broken:

• Progress bar always showed MAX LEVEL! instead of your actual XP progress
• All issue counts, test counts, and doc counts showed 0
• Smart merge during updates was not correctly preserving your XP data

v1.3.1

[1.3.1] - 2026-03-10

Bug Fix

Version display was broken after checking for updates

When you ran npx ocs-stats update and everything was already up to date, the output showed:

Already up to date (vundefined)

This is now fixed. It will correctly show:

Already up to date (v1.3.1)

v1.3.0

[1.3.0] - 2026-03-10

Smarter, Safer Agents

A big round of improvements to make all agents more consistent, more predictable, and harder to misuse.

---

Agents now work in two clear steps

Before today, agents could award themselves points just for reviewing or planning — even if nothing was actually fixed. That's now gone.

Every agent follows a strict two-step process:

• Step 1 — Look & Plan: The agent reads your code, finds issues, and presents them to you. No points awarded.
• Step 2 — Fix: Only after you ask the agent to fix something, and the fix is done and verified, does it earn points.

This means points now reflect real work — not just analysis.

---

Agents remember your project preferences

Each agent now keeps a memory file that grows over time. When you tell an agent "don't use X" or "avoid Y in this project", it writes that down and remembers it in every future session.

Example: if you don't want useEffect used for authentication, the agent records that and will always look for a better approach first — only falling back to useEffect if there is truly no other option, and even then it will ask you first.

---

Two things agents will never do by default

All agents now share two hard rules that apply everywhere:

• No any types — agents will never write or suggest TypeScript's any type. If there's genuinely no other way, they'll flag it to you before doing anything.
• No useEffect by default — agents treat useEffect as a last resort. They'll always try to find a cleaner approach first.

---

Agents build on their skills as they level up

Previously, a Level 3 agent might only focus on Level 3 tasks and ignore the basics. Now every agent applies everything it has learned from Level 1 up to its current level — nothing gets skipped.

---

Level progression is now per-level, not cumulative

XP now resets at each level. When you level up, any overflow carries into the next level.

Level	Points needed to advance
1 - Novice	150
2 - Apprentice	300
3 - Practitioner	450
4 - Expert	1,500
5 - Master	3,000
6 - Grandmaster	Max level

---

Agents can never commit or push code

All agents now have a hard block — they cannot run git commit, git push, or any other git command that changes your repository history. Ever. This rule cannot be overridden.

v1.2.6

[1.2.6] - 2026-02-28

Improved Update Experience

No more conflicts when updating!

Previously, running npx ocs-stats update could show conflicts if you customized your agent files. Now, updates are smart merged:

• Your customizations are preserved (mode, tools, model, permissions, etc.)
• Agent instructions are updated to the latest version
• No manual conflict resolution needed

What you'll see now:

Smart Merged (10):
  ◊ agents/testing.md (mode, tools preserved)
  ◊ agents/security.md (updated)
  ... and more

v1.2.5

[1.2.5] - 2026-02-28

Added

Stats Display for All Agents

• code-review stats - Shows total reviews, issues found by severity
• docs stats - Shows total docs, sections, improvements, code examples, tutorials

Simplified CLI Commands

• npx ocs-stats security - Show security agent progress (was: npx ocs-stats stats)
• npx ocs-stats testing - Show testing agent progress (was: npx ocs-stats stats testing)
• npx ocs-stats code-review - Show code-review agent progress
• npx ocs-stats docs - Show docs agent progress

Updated display-xp Command

• Now supports all 4 agents via [agent] tag
• npx ocs-stats display-xp 35 "Fixed high issue [security]"
• npx ocs-stats display-xp 80 "Wrote 8 tests [testing]"
• npx ocs-stats display-xp 25 "Reviewed PR [code-review]"
• npx ocs-stats display-xp 20 "Updated docs [docs]"

Changed

• bin/cli.js - Agent names now direct commands (removed stats subcommand)
• src/display.js - Added CATEGORY_CONFIG mapping for all 4 agents
• src/init.js - Updated installation message with all agents and new commands

v1.2.4

[1.2.4] - 2026-02-28

Added

New Agents

• code-review agent - Primary mode agent for analyzing code quality, identifying issues, and suggesting improvements

  • Read-only tools (no write/edit/bash)
  • XP system: Review file (+10), catch bug (+25), find security issue (+30)
  • Level progression from Novice to Grandmaster
  • Focus areas: code quality, patterns, best practices, performance, security

• docs agent - Primary mode agent for writing and maintaining project documentation

  • Write/edit tools enabled (no bash)
  • XP system: Write section (+20), create tutorial (+40), add examples (+10)
  • Level progression from Novice to Grandmaster
  • Focus areas: README, API docs, tutorials, code comments, architecture docs

v1.2.3

What's Changed

• chore:bump ver by @Zen0space in #12

Full Changelog: v1.2.2...v1.2.3