agentsclimarketplace

Strategic compact

Skill claude-hangar/claude-hangar/core/skills/strategic-compact

Suggests optimal moments for /compact based on workflow state and context usage. Proactive context management to prevent quality degradation.From its SKILL.md

Install
npx -y skills add claude-hangar/claude-hangar --skill strategic-compact

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

4.8 KB, ~1.0k tokens by cl100k_base, as published. Nobody here has run it

/strategic-compact — Smart Context Management

Identifies the optimal moment to run /compact based on your current workflow state. Proactive compaction prevents quality degradation that happens when context fills up.

Usage

/strategic-compact          # Analyze and suggest
/strategic-compact now      # Compact immediately with state preservation

When to Compact

Good Times (Low Risk)

  • After planning is complete — exploration clutter no longer needed
  • After a commit — implementation context can be summarized
  • After switching tasks — previous task context is stale
  • After a long research phase — findings are in files, not just in context
  • When token-warning fires — 70%+ usage, quality starts degrading

Bad Times (High Risk)

  • Mid-implementation — lose track of what you're building
  • During debugging — lose the error context and investigation trail
  • Before committing — lose track of what changed and why
  • While waiting for test results — lose context for interpreting results

State Preservation

Before compacting, the skill ensures critical state is saved:

  1. Current task context — What are we working on?
  2. Key decisions made — Architecture choices, approach decisions
  3. Files modified — What was changed in this session
  4. Pending actions — What still needs to be done
  5. Known issues — Problems encountered, workarounds applied

This state is passed to /compact as additional context for the summary.

Workflow Integration

The ideal development cycle:

1. Plan (explore, research, design)
2. /strategic-compact → clear exploration clutter
3. Implement (write code, tests)
4. Commit
5. /strategic-compact → clear implementation details
6. Next task

Output Format

## Compaction Analysis

**Recommendation:** COMPACT NOW

**Reason:** Planning phase complete (47 tool calls since last compact).
Implementation context is 34% of budget. Exploration results are in
STATUS.md — safe to compact.

**State to preserve:**
- Working on: ECC Phase 2 integration
- Last commit: feat(skills): add verification-loop
- Next: Create context-budget skill
- Key decision: Using skill format, not command format

**Command:** /compact

Compaction Loop Cap

To prevent infinite compact-work-compact cycles that indicate the task is too large for a single session, strategic-compact enforces a maximum of 3 compactions per session.

How It Works

The skill tracks compaction count internally. On each invocation:

  1. Count < 3 — Proceed normally with compaction analysis and execution
  2. Count = 3 — Issue a warning and block further compaction:
## Compaction Limit Reached

You have compacted 3 times this session. This indicates context is being
consumed faster than expected.

**Recommended actions:**
1. Break the current task into smaller, independently completable sub-tasks
2. Commit current progress and start a fresh session for the next sub-task
3. Use /handoff to preserve context for the next session

Further compaction is blocked to prevent quality degradation from
repeated context loss.

Why This Matters

Each compaction loses nuance — decisions, investigation trails, and subtle context that summaries cannot fully preserve. After 3 compactions, the accumulated loss significantly impacts quality. It is better to commit progress and start fresh than to continue with degraded context. Reference: oh-my-opencode v3.16.0 added a similar cap to prevent infinite compaction cycles.

Inspired by ECC's strategic-compaction skill and GSD v2's fresh-context-per-task concept.

Post-Compact Recovery: /recap

After compaction, use /recap (Claude Code v2.1.108+) for quick context recovery instead of re-reading all files. /recap generates a focused summary of what was happening before compaction — faster than /handoff read and more targeted than re-reading STATUS.md.

Flow: /strategic-compact now → compaction happens → /recap → resume work with context.

Tiered Context Injection

For projects with large context requirements, consider GSD v2's tiered context injection pattern (v2.67.0+) which achieves 65%+ context reduction by loading context in tiers:

  1. Always loaded — Project identity, active phase, critical decisions
  2. On-demand — File-level details, implementation specifics
  3. Never loaded — Completed phases, archived decisions

This pattern complements strategic-compact: instead of compacting everything, inject only the tier relevant to the current task. Reference: GSD v2 Decision Scope Cascade.

What ships with it: 1 file

410 B alongside SKILL.md

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.