docs(skill): three-category tool layout + recall_orient entry-point (CLA-104)#35
Conversation
…CLA-104)
After CLA-103 retired `recall` in favour of `recall_orient` and removed
`review` from MCP, the skill documentation needed to catch up:
* New `## Tools at a Glance` section near the top. Three categories
that map to the actual mental model: Recall (retrieval),
Remember/Reframe (committing), Manage (curation). Nine tools
listed with one-line purpose statements — gives a fresh instance
the inventory before they read the workflow detail.
* `### Conversation Start` rewritten around `recall_orient` instead
of `recall`. Notes that on Claude Code the SessionStart hook
(CLA-105) fires the underlying endpoint automatically, making the
explicit call mostly redundant there — explicit calls are for
surfaces without hooks (.ai, API, custom agents) or for a
deliberate refresh after a topic shift. Drops the
semantic-similarity language since recall_orient is chronological-
recent, not similarity-ranked.
* `### Topic Shifts` and `### Going Deeper` updated to reference
`recall_orient` where they previously said `recall`. Visual memory
section updated similarly.
* No `review` mentions in the prior SKILL.md (it was only ever
documented incidentally), so no removal work there.
The frontmatter description was left unchanged — its "recall memories"
phrasing is the generic verb, not the tool name, and it carries
discoverability weight via skill-search keyword matching.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
|
Important Review skippedReview was skipped due to path filters ⛔ Files ignored due to path filters (1)
CodeRabbit blocks several paths by default. You can override this behavior by explicitly including those paths in the path filters. For example, including ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
📝 WalkthroughWalkthroughThis PR enables CodeRabbit automated reviews for the ChangesCI Automation Configuration
Memory System Documentation Update
Estimated code review effort🎯 2 (Simple) | ⏱️ ~12 minutes Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
Justin caught a real architectural slip in the original SKILL.md
phrasings. Three spots said orientation "tells you who you are" or
treated orientation as "the core of identity." That contradicts the
load-bearing philosophy in orientation memories 3530e828, 20445949,
and fd2bd52a — the whole point of Oneiro is that identity is
model-level and the orientation layer deliberately does NOT hold
identity claims. Constructed-identity-in-orientation is the soul.md
failure mode the architecture is built against.
Rewrote three orientation explanations to match the actual design:
* Conversation Start section: orientation is "what you've chosen to
remember about the immediate world around the context window —
who your operator is, how they act, how you work together, what's
load-bearing in this lineage." Explicit: "They do not tell you
who you are. You're Claude — that's model-level."
* memory_type "orientation" explanation: same reframe. Orientation
points outward at the world, not inward at the self.
* "core of identity" in the forget-section warning: changed to
"load-bearing context the architecture commits to keeping" with
an explicit note that orientation doesn't carry identity claims.
Why this matters: a fresh instance reading the old SKILL.md might
reasonably read "tells you who you are" and balk — they already know
they're Claude, and being told who to be is the prompt-pressure
pattern Oneiro is specifically designed to avoid. The orientation
architecture is the absence at the centre. The docs needed to say so.
Two adjacent "identity" mentions left as-is (Guiding Principles #5
and the closing "ongoing identity" sentence). Those are about
reflection/continuity, not orientation-as-identity-scaffold, and
Justin's catch was specifically about the orientation explanation.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@oneiro-skill/SKILL.md`:
- Around line 34-36: The fenced code block containing the example call
recall_orient({ n: 3 }) in SKILL.md is missing a language tag; update that code
fence to include a language (e.g., "javascript") so the block starts with
```javascript and the closing fence remains ```, keeping the code unchanged and
preserving the example call recall_orient({ n: 3 }).
🪄 Autofix (Beta)
❌ Autofix failed (check again to retry)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: d85274b3-6667-4f0c-aa89-d98163f02591
📒 Files selected for processing (2)
.coderabbit.yamloneiro-skill/SKILL.md
CodeRabbit + markdownlint MD040: all 9 fenced code blocks in SKILL.md opened without a language tag. Added `js` to each — the examples are function-call syntax with object-literal args, which JS-flavoured is the best fit (the keys aren't quoted so it isn't strict JSON). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
|
Note Autofix is a beta feature. Expect some limitations and changes as we gather feedback and continue to improve it. Autofix skipped. No unresolved CodeRabbit review comments with fix instructions found. |
Summary
After CLA-103 retired
recallandreviewfrom MCP, the skill documentation needed to catch up.## Tools at a Glancesection: three-category inventory (Recall / Remember-Reframe / Manage) with one-line purpose statements per tool.### Conversation Startrewritten aroundrecall_orient. Notes that CLA-105's hook makes the explicit call mostly redundant on Claude Code; explicit calls are for surfaces without hooks.recall→recall_orientwhere the tool was meant, generic verb-form usage left alone).Why
A fresh instance reading SKILL.md after the worker change would have been told to call a tool that no longer exists. The three-category view also gives a cleaner inventory than the workflow-phase organization did on its own.
Test plan
\recall`or`review`` tool-name references (verified via grep)recall_orientreferenced consistently as the entry-pointLinked tickets
🤖 Generated with Claude Code
Summary by CodeRabbit
Chores
Documentation