Support prefix matching for shortened thought IDs

[timeleft--]

Problem

Thought IDs are 26-character ULIDs (e.g. 01KKYAACFAYFH7F727HVTNE9T3). Agents routinely shorten these to 8-character prefixes (e.g. 01KKYAAC) in their tool calls — this is natural behavior since agents display thoughts in tables with shortened IDs.

Currently, all thought-by-ID operations (get_thought, update_thought, supersede, propose_truth, change_scope) require an exact full ULID match via _find_thought_path(). When an agent passes a shortened ID, the operation fails with "not found".

Expected Behavior

• If a shortened ID prefix uniquely matches exactly one thought file, resolve it and proceed normally
• If a prefix matches multiple thought files, return an error listing all matching thought IDs with their YAML frontmatter metadata (source_type, created_at, content_preview, etc.) so the agent can disambiguate
• Exact full-ULID matches continue to work as before (fast path)

Example

Agent displays thoughts in a table:

│  5   │ 01KKYAAC │ .gitignore template for builder worktree compatibility │ decision    │

Then calls:

propose_truth(trail_name="mwai/eng/codev", thought_id="01KKYAAC")

This should resolve 01KKYAAC → 01KKYAACFAYFH7F727HVTNE9T3 and promote it.

Implementation Notes

• Core change is in TrailManager._find_thought_path() (trail.py:104-109)
• Add an AmbiguousThoughtID exception class with structured candidate info
• Tool handlers in tools/thought.py and tools/navigation.py should catch AmbiguousThoughtID and return a helpful error response
• Add tests for: unique prefix match, ambiguous prefix match, exact match still works

Labels

bug

[timeleft--]

On it! Working on a fix now.

[timeleft--]

Resolved in PR #33 (commits a83abb0, 8d72559) — prefix matching for shortened thought IDs is now supported in _find_thought_path().