claude-praetorian

name	claude-praetorian
description	Cross-session memory - save insights after research, restore context at session start. Reduces re-exploration and prevents token waste.
triggers	["PreToolUse","PostToolUse","SubagentStop","PreCompact"]

Praetorian Plugin

Context guard with automatic workflow nudges. Hooks fire at natural decision points — you decide whether to act on them.

How Hooks Work

Hooks emit one-line nudges (~20 tokens each). No disk I/O. You judge relevance.

Hook	When It Fires	What It Says	Your Call
PreToolUse	Before Task, WebSearch, WebFetch	"Check restore — prior context may exist"	Restore if topic is familiar. Skip if clearly novel.
PostToolUse	After WebFetch, WebSearch	"Save key findings"	Compact research results. Reuse title to merge.
SubagentStop	After subagent completes	"Save key findings"	Compact subagent results. Don't let findings die with the agent.
PreCompact	Before any compact call	"Reuse title to merge"	Always follow — prevents duplicate compactions.

Reading Hook Nudges

When you see ⚜️ Check praetorian_restore(query="..."):

1. Is this a topic you've worked on before? → Restore (summary mode first)
2. Is the summary hit relevant? → Restore again with detail="full"
3. Clearly novel topic? → Skip, proceed with research

When you see ⚜️ Save key findings: praetorian_compact(...):

1. Compact with existing title to merge — don't wait for "enough" findings
2. Routine result, nothing novel? → Skip

Workflow: restore → work → compact

Phase 1: Restore (Before Work)

praetorian_restore(query="auth JWT")                    # summary scan (~50 tokens/result)
praetorian_restore(query="auth JWT", detail="full")     # full detail for relevant hit
praetorian_restore(project="current")                   # scope to this project
praetorian_restore()                                    # recent compactions

Hooks handle: PreToolUse nudges before Task/WebSearch/WebFetch. You handle manually: Before planning sessions, before launching subagents, at session start, after /clear.

Phase 2: Work

Do your research, run subagents, make decisions. Hooks will nudge after.

Phase 3: Compact (After Valuable Findings)

praetorian_compact(type="web_research", title="auth JWT research",
  key_insights=["JWT rotation every 15min", "RS256 preferred over HS256"],
  refs=["src/auth/middleware.ts:45"])

Hooks handle: PostToolUse nudges after WebSearch/WebFetch, SubagentStop nudges after agents. You handle manually: After subagents return, after multi-file exploration, after solving a hard problem. Don't wait — compact incrementally.

Critical: Always reuse an existing title to merge. Creating a new compaction costs 1 of 13 slots per project. Merging is free.

Phase 4: Manage (When Needed)

praetorian_manage(action="status")              # per-project counts, disk usage
praetorian_manage(action="prune", count=3)      # remove 3 oldest
praetorian_manage(action="prune", project="/path/to/old-project")

Auto-limit: 13 per project. Oldest auto-pruned when exceeded. Only surface to user when >50 total or >500KB. If store is healthy, stay quiet.

Sibling Plugin Synergy

When historian is active, check historian FIRST for conversation-level context, then praetorian for distilled insights. They complement:

• Historian = raw session transcripts, broad search
• Praetorian = curated insights, structured and compact

When vigil is active, vigil handles file checkpoints. Praetorian handles knowledge checkpoints. Don't duplicate — vigil saves files, praetorian saves understanding.

Compaction Quality

Good compaction:

• Specific title matching the topic (enables auto-merge)
• 3-5 key_insights that save future research time
• File:line refs that anchor claims to code
• Decisions with rationale (chose X over Y because Z)

Bad compaction:

• Vague title like "research findings"
• Restating what's obvious from the code
• Raw content dump without distillation
• Creating new when a merge-able compaction exists

Token Discipline

• Summary first: Default restore is ~50 tokens/result. Full restore only for confirmed hits.
• Selective compaction: One well-maintained compaction > five stale ones.
• Skip when irrelevant: Hook nudge about "cooking recipes" when working on auth? Skip.
• Don't flood: 3 compactions/session is typical. 10+ means you're not merging enough.

Commands

Command	Description
/compact-praetorian [type] [title]	Save current context
/restore-praetorian [query]	Load previous context