rrr

Name: rrr
Author: Soul-Brews-Studio

109stars55forks

How to install rrr?

soul-brews-studio/arra-oracle-skills-cli/rrr

$npx -y skills add Soul-Brews-Studio/arra-oracle-skills-cli --skill rrr

Installs into the current project.

Summary

Create session retrospective with AI diary and lessons learned. Use when user says "rrr", "retrospective", "wrap up session", "session summary", or at end of work session.

Files · 1

View on GitHub

SKILL.md

1	# /rrr
2
3	> "Reflect to grow, document to remember."
4
5	```
6	/rrr # Retro + 1 background dig subagent (parallel, fast)
7	/rrr --quick # No dig, no subagent — memory only (fastest)
8	/rrr --detail # Full template + background dig
9	/rrr --deep # 5 parallel subagents
10	/rrr --deep --teammate # 3 coordinated team agents (requires AGENT_TEAMS)
11	```
12
13	Default mode: main agent starts writing the retro immediately from conversation memory.
14	One background subagent runs dig + .jsonl timestamp extraction in parallel.
15	When the subagent returns, main agent merges real timestamps into the Timeline section.
16	No speed penalty — dig runs while you write.
17
18	`--quick` skips dig entirely — memory only, zero subagents.
19
20	Subagent rules: default /rrr spawns exactly 1 background Agent (dig miner). `--deep` spawns 5. `--quick` spawns 0.
21	NEVER use the Task tool. Only `--deep` and `--deep --teammate` use TeamCreate.
22
23	---
24
25	## Oracle Root Detection (REQUIRED — run before any ψ/ write)
26
27	Every skill that writes to ψ/ MUST detect the oracle root first. Do not assume `pwd` is the oracle repo.
28
29	```bash
30	# Step 1: Find git root
31	ORACLE_ROOT=$(git rev-parse --show-toplevel 2>/dev/null)
32
33	# Step 2: Cross-check — oracle repo has CLAUDE.md + ψ/
34	if [ -n "$ORACLE_ROOT" ] && [ -f "$ORACLE_ROOT/CLAUDE.md" ] && { [ -d "$ORACLE_ROOT/ψ" ] \|\| [ -L "$ORACLE_ROOT/ψ" ]; }; then
35	PSI="$ORACLE_ROOT/ψ"
36	elif [ -f "$(pwd)/CLAUDE.md" ] && { [ -d "$(pwd)/ψ" ] \|\| [ -L "$(pwd)/ψ" ]; }; then
37	# Fallback: pwd has oracle markers
38	ORACLE_ROOT="$(pwd)"
39	PSI="$ORACLE_ROOT/ψ"
40	else
41	# Last resort: warn and use pwd
42	echo "⚠️ Not in oracle repo (no CLAUDE.md + ψ/ at git root). Writing to pwd."
43	ORACLE_ROOT="$(pwd)"
44	PSI="$ORACLE_ROOT/ψ"
45	fi
46	```
47
48	Why: prevents retros writing to `~/ψ/` (home) or incubated repo's `ψ/` instead of the oracle's own vault.
49
50	All paths below use `$PSI/` instead of bare `ψ/`.
51
52	---
53
54	## /rrr (Default — background dig + parallel write)
55
56	### 1. Gather git context (main agent)
57
58	```bash
59	date "+%H:%M %Z (%A %d %B %Y)"
60	git log --oneline -10 && git diff --stat HEAD~5
61	```
62
63	Detect session ID:
64
65	```bash
66	ENCODED_PWD=$(echo "$ORACLE_ROOT" \| sed 's\|^/\|-\|; s\|[/.]\|-\|g')
67	PROJECT_BASE=$(ls -d "$HOME/.claude/projects/${ENCODED_PWD}" 2>/dev/null \| head -1)
68	LATEST_JSONL=$(ls -t "$PROJECT_BASE"/*.jsonl 2>/dev/null \| head -1)
69	[ -n "$LATEST_JSONL" ] && SESSION_ID=$(basename "$LATEST_JSONL" .jsonl) && echo "SESSION: ${SESSION_ID:0:8}"
70	```
71
72	### 1.5. Spawn timestamp miner (background subagent)
73
74	Spawn ONE background Agent to extract real timestamps from the session .jsonl:
75
76	```
77	Agent({
78	name: "timestamp-miner",
79	description: "Extract session timestamps for /rrr",
80	run_in_background: true,
81	prompt: `Extract real user message timestamps from a Claude Code session file.
82	Read-only — do NOT write files.
83
84	Run this single command:
85
86	ENCODED_PWD=$(echo "[ORACLE_ROOT]" \| sed 's\|^/\|-\|; s\|[/.]\|-\|g')
87	PROJECT_BASE=$(ls -d "$HOME/.claude/projects/${ENCODED_PWD}" 2>/dev/null \| head -1)
88	LATEST_JSONL=$(ls -t "$PROJECT_BASE"/*.jsonl 2>/dev/null \| head -1)
89	echo "SESSION_FILE: $LATEST_JSONL"
90	python3 -c "
91	import json, os, sys
92	sys.stdout.reconfigure(encoding='utf-8') # Windows Thai locale: stdout defaults to cp874 -> Thai snippets mojibake; force UTF-8 output
93	from datetime import datetime, timezone, timedelta
94	tz = timezone(timedelta(hours=7))
95	jsonl = '$LATEST_JSONL'
96	if not jsonl or not os.path.exists(jsonl): exit(0)
97	with open(jsonl, encoding='utf-8') as f:
98	for line in f:
99	try:
100	m = json.loads(line)
101	if m.get('type') != 'user' or 'message' not in m: continue
102	content = m['message'].get('content', '')
103	if isinstance(content, list):
104	for c in content:
105	if isinstance(c, dict) and c.get('type') == 'text':
106	content = c.get('text', ''); break
107	if not isinstance(content, str): continue
108	ts = m.get('timestamp', '')
109	if not ts or '<command-name>' in content[:200]: continue
110	dt = datetime.fromisoformat(ts.replace('Z', '+00:00')).astimezone(tz)
111	snippet = content[:80].replace(chr(10), ' ')
112	print(f'{dt.strftime(\"%Y-%m-%d %H:%M\")} \| {snippet}')
113	except: pass
114	"
115
116	Return ALL output lines. The main agent will use them for the Timeline.`
117	})
118	```
119
120	Why only .jsonl, not dig.py: the subagent has no conversation context — it can't interpret dig session summaries. The .jsonl timestamps are objective data (real ISO timestamps from every user message). That's all we need for the Timeline.
121
122	### 2. Write Retrospective (main agent — start immediately, don't wait for dig)
123
124	Path: `$PSI/memory/retrospectives/YYYY-MM/DD/

People who liked this also viewed

6 picks

Type

find-skillsHelps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skSkillsJul 20262.6M27k
grill-meA relentless interview to sharpen a plan or design.SkillsJul 2026633k182k
vercel-react-best-practicesVercel's official collection of agent skillsSkillsJul 2026573k29k
improve-codebase-architectureScan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.SkillsJul 2026522k182k
setup-matt-pocock-skillsConfigure this repo for the engineering skills — set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.SkillsJul 2026454k182k
triageMove issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.SkillsJul 2026424k182k

rrr

109stars55forks

View on GitHub

How to install rrr?

soul-brews-studio/arra-oracle-skills-cli/rrr

$npx -y skills add Soul-Brews-Studio/arra-oracle-skills-cli --skill rrr

Installs into the current project.

Summary

Create session retrospective with AI diary and lessons learned. Use when user says "rrr", "retrospective", "wrap up session", "session summary", or at end of work session.

Files · 1

View on GitHub

SKILL.md

1	# /rrr
2
3	> "Reflect to grow, document to remember."
4
5	```
6	/rrr # Retro + 1 background dig subagent (parallel, fast)
7	/rrr --quick # No dig, no subagent — memory only (fastest)
8	/rrr --detail # Full template + background dig
9	/rrr --deep # 5 parallel subagents
10	/rrr --deep --teammate # 3 coordinated team agents (requires AGENT_TEAMS)
11	```
12
13	Default mode: main agent starts writing the retro immediately from conversation memory.
14	One background subagent runs dig + .jsonl timestamp extraction in parallel.
15	When the subagent returns, main agent merges real timestamps into the Timeline section.
16	No speed penalty — dig runs while you write.
17
18	`--quick` skips dig entirely — memory only, zero subagents.
19
20	Subagent rules: default /rrr spawns exactly 1 background Agent (dig miner). `--deep` spawns 5. `--quick` spawns 0.
21	NEVER use the Task tool. Only `--deep` and `--deep --teammate` use TeamCreate.
22
23	---
24
25	## Oracle Root Detection (REQUIRED — run before any ψ/ write)
26
27	Every skill that writes to ψ/ MUST detect the oracle root first. Do not assume `pwd` is the oracle repo.
28
29	```bash
30	# Step 1: Find git root
31	ORACLE_ROOT=$(git rev-parse --show-toplevel 2>/dev/null)
32
33	# Step 2: Cross-check — oracle repo has CLAUDE.md + ψ/
34	if [ -n "$ORACLE_ROOT" ] && [ -f "$ORACLE_ROOT/CLAUDE.md" ] && { [ -d "$ORACLE_ROOT/ψ" ] \|\| [ -L "$ORACLE_ROOT/ψ" ]; }; then
35	PSI="$ORACLE_ROOT/ψ"
36	elif [ -f "$(pwd)/CLAUDE.md" ] && { [ -d "$(pwd)/ψ" ] \|\| [ -L "$(pwd)/ψ" ]; }; then
37	# Fallback: pwd has oracle markers
38	ORACLE_ROOT="$(pwd)"
39	PSI="$ORACLE_ROOT/ψ"
40	else
41	# Last resort: warn and use pwd
42	echo "⚠️ Not in oracle repo (no CLAUDE.md + ψ/ at git root). Writing to pwd."
43	ORACLE_ROOT="$(pwd)"
44	PSI="$ORACLE_ROOT/ψ"
45	fi
46	```
47
48	Why: prevents retros writing to `~/ψ/` (home) or incubated repo's `ψ/` instead of the oracle's own vault.
49
50	All paths below use `$PSI/` instead of bare `ψ/`.
51
52	---
53
54	## /rrr (Default — background dig + parallel write)
55
56	### 1. Gather git context (main agent)
57
58	```bash
59	date "+%H:%M %Z (%A %d %B %Y)"
60	git log --oneline -10 && git diff --stat HEAD~5
61	```
62
63	Detect session ID:
64
65	```bash
66	ENCODED_PWD=$(echo "$ORACLE_ROOT" \| sed 's\|^/\|-\|; s\|[/.]\|-\|g')
67	PROJECT_BASE=$(ls -d "$HOME/.claude/projects/${ENCODED_PWD}" 2>/dev/null \| head -1)
68	LATEST_JSONL=$(ls -t "$PROJECT_BASE"/*.jsonl 2>/dev/null \| head -1)
69	[ -n "$LATEST_JSONL" ] && SESSION_ID=$(basename "$LATEST_JSONL" .jsonl) && echo "SESSION: ${SESSION_ID:0:8}"
70	```
71
72	### 1.5. Spawn timestamp miner (background subagent)
73
74	Spawn ONE background Agent to extract real timestamps from the session .jsonl:
75
76	```
77	Agent({
78	name: "timestamp-miner",
79	description: "Extract session timestamps for /rrr",
80	run_in_background: true,
81	prompt: `Extract real user message timestamps from a Claude Code session file.
82	Read-only — do NOT write files.
83
84	Run this single command:
85
86	ENCODED_PWD=$(echo "[ORACLE_ROOT]" \| sed 's\|^/\|-\|; s\|[/.]\|-\|g')
87	PROJECT_BASE=$(ls -d "$HOME/.claude/projects/${ENCODED_PWD}" 2>/dev/null \| head -1)
88	LATEST_JSONL=$(ls -t "$PROJECT_BASE"/*.jsonl 2>/dev/null \| head -1)
89	echo "SESSION_FILE: $LATEST_JSONL"
90	python3 -c "
91	import json, os, sys
92	sys.stdout.reconfigure(encoding='utf-8') # Windows Thai locale: stdout defaults to cp874 -> Thai snippets mojibake; force UTF-8 output
93	from datetime import datetime, timezone, timedelta
94	tz = timezone(timedelta(hours=7))
95	jsonl = '$LATEST_JSONL'
96	if not jsonl or not os.path.exists(jsonl): exit(0)
97	with open(jsonl, encoding='utf-8') as f:
98	for line in f:
99	try:
100	m = json.loads(line)
101	if m.get('type') != 'user' or 'message' not in m: continue
102	content = m['message'].get('content', '')
103	if isinstance(content, list):
104	for c in content:
105	if isinstance(c, dict) and c.get('type') == 'text':
106	content = c.get('text', ''); break
107	if not isinstance(content, str): continue
108	ts = m.get('timestamp', '')
109	if not ts or '<command-name>' in content[:200]: continue
110	dt = datetime.fromisoformat(ts.replace('Z', '+00:00')).astimezone(tz)
111	snippet = content[:80].replace(chr(10), ' ')
112	print(f'{dt.strftime(\"%Y-%m-%d %H:%M\")} \| {snippet}')
113	except: pass
114	"
115
116	Return ALL output lines. The main agent will use them for the Timeline.`
117	})
118	```
119
120	Why only .jsonl, not dig.py: the subagent has no conversation context — it can't interpret dig session summaries. The .jsonl timestamps are objective data (real ISO timestamps from every user message). That's all we need for the Timeline.
121
122	### 2. Write Retrospective (main agent — start immediately, don't wait for dig)
123
124	Path: `$PSI/memory/retrospectives/YYYY-MM/DD/

People who liked this also viewed

6 picks

Type