claude-compaction-restore

Name: claude-compaction-restore
Author: mvschwarz

56stars9forks

How to install claude-compaction-restore?

mvschwarz/openrig/claude-compaction-restore

$npx -y skills add mvschwarz/openrig --skill claude-compaction-restore

Installs into the current project.

Summary

Use when a Claude Code session has just compacted, is about to compact, reached context limit, resumed after /compact, or needs to rebuild its working mental model from Claude JSONL transcripts and touched files.

Files · 1

View on GitHub

SKILL.md

1	# Claude Compaction Restore
2
3	Use this skill to preserve continuity before Claude Code compacts and to
4	restore continuity after compaction. Do the work described here; do not merely
5	explain the protocol.
6
7	## If You Are About To Compact
8
9	Prepare durable continuity before the context boundary.
10
11	1. Identify the active task, queue item, mission/slice, branch or commit, and
12	current working directory.
13	2. Record the current state: decisions made, files changed, commands/tests run,
14	evidence produced, blockers, caveats, and the next concrete step.
15	3. Create or update a durable mental-model restore map. This map is the main
16	artifact future-you will use to rebuild context after compaction.
17	4. In the restore map, write an ASCII file/folder tree of every path that
18	mattered to your working mental model during this session. Include:
19	- the active queue item or mission packet;
20	- mission notes, progress, decisions, and evidence files;
21	- Claude memory/project notes you used or wrote, especially when the memory
22	folder is shared by many agents;
23	- files with active edits or recently inspected source;
24	- root instructions such as `AGENTS.md`, `CLAUDE.md`, or `README.md`;
25	- as-built docs, codemaps, conventions, skills, and product docs needed
26	before code/review work;
27	- source files, tests, scripts, UI evidence, screenshots, logs, or reports
28	that shaped your current state.
29	5. For every file or folder in the tree, add a short note explaining why it
30	matters and whether it is required reading after compaction.
31	6. Write any important glue context that is not already on disk into the
32	handoff/restore map. This includes assumptions, partial conclusions, failed
33	paths, and why the listed files fit together.
34	7. In the compaction summary, include the restore map path and the top required
35	reading paths from that map.
36
37	## If You Just Compacted
38
39	Treat your memory as unreliable until you restore from durable evidence.
40	Restoration is your current task. Do not wait for a future user request or a
41	new task assignment before reading the required files.
42
43	1. Read any restore prompt, marker path, packet path, transcript path, restore
44	map, or extra instruction file named by the operator or OpenRig.
45	2. Run the restore script from this skill:
46
47	```bash
48	node ~/.claude/skills/claude-compaction-restore/scripts/restore-from-jsonl.mjs --out /tmp/claude-compaction-restore
49	```
50
51	If the prompt names a specific Claude JSONL transcript, pass it explicitly:
52
53	```bash
54	node ~/.claude/skills/claude-compaction-restore/scripts/restore-from-jsonl.mjs /path/to/session.jsonl --out /tmp/claude-compaction-restore
55	```
56
57	3. Read the generated `restore-instructions.md`.
58	4. Read the generated `touched-files.md`.
59	5. Find and read the mental-model restore map if one exists. Treat it as the
60	highest-value guide to which memory files, mission files, source files, and
61	docs mattered before compaction.
62	6. Identify the important files for the active task. The touched-file list is
63	a triage aid, not a complete inventory.
64	7. Read every important file in full. Prioritize the restore map, Claude memory
65	files named in the map, Markdown state/planning files, queue/mission packets,
66	source files with active edits, root instruction files, and as-built or
67	codemap docs.
68	8. After the required reads are complete, state:
69
70	```text
71	restored from packet at <path>; resumed at step <X>
72	```
73
74	Include the main files you read in full when you make that statement.
75
76	## Required Read-Depth Audit
77
78	After the first restore pass, audit yourself before continuing.
79
80	1. List every file, packet, marker, restore map, instruction file, and source
81	document you were asked to read during restore.
82	2. Mark each item as `FULL`, `PARTIAL`, or `NOT_READ`.
83	3. You will be given a task where all of these files are required reading in
84	order to understand the task.
85	4. Do not optimize for token conservation.
86	5. Read every `PARTIAL` or `NOT_READ` item in full now.
87	6. Report the final read-depth table before doing any substantive task work.
88
89	## Guardrails
90
91	- Do not silently launch fresh after compaction.
92	- Do not continue from memory when restore evidence exists.
93	- Do not defer required restore reading until a later user task. The restore is
94	t

People who liked this also viewed

6 picks

Type

claude-compaction-restore

56stars9forks

View on GitHub

How to install claude-compaction-restore?

mvschwarz/openrig/claude-compaction-restore

$npx -y skills add mvschwarz/openrig --skill claude-compaction-restore

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Claude Compaction Restore
2
3	Use this skill to preserve continuity before Claude Code compacts and to
4	restore continuity after compaction. Do the work described here; do not merely
5	explain the protocol.
6
7	## If You Are About To Compact
8
9	Prepare durable continuity before the context boundary.
10
11	1. Identify the active task, queue item, mission/slice, branch or commit, and
12	current working directory.
13	2. Record the current state: decisions made, files changed, commands/tests run,
14	evidence produced, blockers, caveats, and the next concrete step.
15	3. Create or update a durable mental-model restore map. This map is the main
16	artifact future-you will use to rebuild context after compaction.
17	4. In the restore map, write an ASCII file/folder tree of every path that
18	mattered to your working mental model during this session. Include:
19	- the active queue item or mission packet;
20	- mission notes, progress, decisions, and evidence files;
21	- Claude memory/project notes you used or wrote, especially when the memory
22	folder is shared by many agents;
23	- files with active edits or recently inspected source;
24	- root instructions such as `AGENTS.md`, `CLAUDE.md`, or `README.md`;
25	- as-built docs, codemaps, conventions, skills, and product docs needed
26	before code/review work;
27	- source files, tests, scripts, UI evidence, screenshots, logs, or reports
28	that shaped your current state.
29	5. For every file or folder in the tree, add a short note explaining why it
30	matters and whether it is required reading after compaction.
31	6. Write any important glue context that is not already on disk into the
32	handoff/restore map. This includes assumptions, partial conclusions, failed
33	paths, and why the listed files fit together.
34	7. In the compaction summary, include the restore map path and the top required
35	reading paths from that map.
36
37	## If You Just Compacted
38
39	Treat your memory as unreliable until you restore from durable evidence.
40	Restoration is your current task. Do not wait for a future user request or a
41	new task assignment before reading the required files.
42
43	1. Read any restore prompt, marker path, packet path, transcript path, restore
44	map, or extra instruction file named by the operator or OpenRig.
45	2. Run the restore script from this skill:
46
47	```bash
48	node ~/.claude/skills/claude-compaction-restore/scripts/restore-from-jsonl.mjs --out /tmp/claude-compaction-restore
49	```
50
51	If the prompt names a specific Claude JSONL transcript, pass it explicitly:
52
53	```bash
54	node ~/.claude/skills/claude-compaction-restore/scripts/restore-from-jsonl.mjs /path/to/session.jsonl --out /tmp/claude-compaction-restore
55	```
56
57	3. Read the generated `restore-instructions.md`.
58	4. Read the generated `touched-files.md`.
59	5. Find and read the mental-model restore map if one exists. Treat it as the
60	highest-value guide to which memory files, mission files, source files, and
61	docs mattered before compaction.
62	6. Identify the important files for the active task. The touched-file list is
63	a triage aid, not a complete inventory.
64	7. Read every important file in full. Prioritize the restore map, Claude memory
65	files named in the map, Markdown state/planning files, queue/mission packets,
66	source files with active edits, root instruction files, and as-built or
67	codemap docs.
68	8. After the required reads are complete, state:
69
70	```text
71	restored from packet at <path>; resumed at step <X>
72	```
73
74	Include the main files you read in full when you make that statement.
75
76	## Required Read-Depth Audit
77
78	After the first restore pass, audit yourself before continuing.
79
80	1. List every file, packet, marker, restore map, instruction file, and source
81	document you were asked to read during restore.
82	2. Mark each item as `FULL`, `PARTIAL`, or `NOT_READ`.
83	3. You will be given a task where all of these files are required reading in
84	order to understand the task.
85	4. Do not optimize for token conservation.
86	5. Read every `PARTIAL` or `NOT_READ` item in full now.
87	6. Report the final read-depth table before doing any substantive task work.
88
89	## Guardrails
90
91	- Do not silently launch fresh after compaction.
92	- Do not continue from memory when restore evidence exists.
93	- Do not defer required restore reading until a later user task. The restore is
94	t

People who liked this also viewed

6 picks

Type