find-session
Skill316 starsChanged 6 months agoMITView on GitHub
Identify a specific Claude Code conversation by its UUID and optionally pin it as a shell function for one-word resume. Use when "claude --continue" picks the wrong session, or when the user wants to reliably resume a specific past conversation across reboots.
We split this skill into the rules it gives an agent
7 rules. 0 are common, 6 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
6
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 28
If unsure which encoded-cwd to scan:
- Distinctive: line 48
Ask: "What's something only this conversation would contain?
- Distinctive: line 89
After write, tell user to either:
- Very specific: line 90
Run source ~/.zshrc in current terminal, OR
- Distinctive: line 99
If user has multiple long-running parallel sessions (e.g., two different projects, both started from HOME), each needs its own pin.
- Distinctive: line 102
Don't confuse session UUIDs across different conversations.
- Distinctive: line 114
Should show the function definition.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
find-session — pin a specific Claude conversation
claude --continue picks "most recent in cwd" — unreliable when multiple
sessions share a directory. claude --resume <uuid> is exact. This skill
finds the right UUID by grepping jsonl files, then writes a one-word shell
function so the user types e.g. aichain to resume that exact conversation.
When to use
- User says "I can't find my session" / "claude --continue picks wrong one"
- User wants a one-word resume command for a specific long-running conversation
- Multiple
.jsonlfiles in~/.claude/projects/<encoded-cwd>/exist
Workflow
1. Find the encoded-cwd directory
Sessions are stored under ~/.claude/projects/<encoded-cwd>/. The encoding
replaces / with -, so cwd /Users/foo/Desktop/MyProj becomes
-Users-foo-Desktop-MyProj. Often the cwd at session-start was the user's
HOME, even if work happened in a subdirectory.
If unsure which encoded-cwd to scan:
ls -dt ~/.claude/projects/*/ | head -5The most recently modified directory is likely the right one.
2. List active session files
ls -lt ~/.claude/projects/<encoded-cwd>/*.jsonl 2>/dev/null | head -10Multiple files = multiple sessions. The biggest by size is NOT necessarily the right one — both the current assistant and other parallel sessions can be wrong about that.
3. Get a unique-content hint from the user
Ask: "What's something only this conversation would contain? A specific filename you created, a unique tool you ran, a project name, etc."
Good hints (highly unique):
- A hostname / URL the conversation generated (e.g.
waqvu5.png) - A file path that was created by this session (e.g.
kimi_research_results) - A specific function name written in this session (e.g.
compute_factor_scores) - A unique error message they remember
Bad hints (too common, will match multiple):
- Generic words like "Python", "test", "fix"
- Common library names
4. grep to find matching session
for f in ~/.claude/projects/<encoded-cwd>/*.jsonl; do
echo "=== $(basename "$f") ==="
grep -c "<unique-string>" "$f" 2>/dev/null
doneThe session with non-zero hits (ideally many hits) is the answer. Confirm with 2-3 different unique strings if the first match is ambiguous.
5. Offer to pin as shell function
Once UUID is identified, ask user for a short name (e.g. aichain,
work, headsup). Then append to ~/.zshrc:
# <project description> — resume specific session
<short-name>() {
cd "$HOME" && claude --resume "<uuid>"
}cd "$HOME" matters: sessions are indexed by their starting cwd, so the
alias must cd to the same place where the session was originally created
(often HOME, even if the actual work touched subdirectories).
After write, tell user to either:
- Run
source ~/.zshrcin current terminal, OR - Open a new terminal
Important notes
- Each
.jsonlis the full transcript log of one session. Sizes can mislead; always rely on grep'd content matches, not file size. - The UUID is stable forever — once pinned, the alias keeps working across reboots, OS upgrades, even years (assuming the .jsonl isn't deleted).
- If user has multiple long-running parallel sessions (e.g., two different projects, both started from HOME), each needs its own pin. Repeat the workflow once per session.
- Don't confuse session UUIDs across different conversations. If two Claude instances both claim the same UUID, at least one is wrong — only grep on actual unique content can verify.
Quick check
After pinning, verify:
source ~/.zshrc && type <short-name>Should show the function definition.
Other versions9 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 9
- kauffj/agent-config★ 8same name
- ooooooooooooooooooop/agent-tools★ 4same name
- ooooooooooooooooooop/personal-ai★ 4same name
- theyoungastronauts/polaris★ 4same name
- FooJiaYin/dev-skills★ 2same name
- LiozShor/claude-code-skills★ 2same name
- wardawgmalvicious/agent-config★ 1same name
- Aman-Kothari7/claude-code-skills★ 0same name
- rhvdbergh/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository3 other files it gives its agents.
- mckinsey-consultantSkill
- mimeng-writingSkill
- us-gov-shutdown-trackerSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

