discord-soul

Name: discord-soul
Author: berabuddies

555stars59forks

How to install discord-soul?

berabuddies/semia/discord-soul

$npx -y skills add berabuddies/Semia --skill discord-soul

Installs into the current project.

Summary

Create a living agent from your Discord server. The agent embodies your community's identity, remembers every conversation, and grows as the community evolves. Talk to your Discord as if it were a person.

Files · 1

View on GitHub

SKILL.md

1	# Discord Soul
2
3	Turn your Discord server into a living, breathing agent.
4
5	## What You Get
6
7	An agent that:
8	- Remembers every conversation in your Discord
9	- Speaks in your community's voice
10	- Knows the key figures, channels, and inside jokes
11	- Grows as new messages arrive daily
12	- Answers questions about your community's history and culture
13
14	## Quick Start
15
16	```bash
17	# Create agent from your Discord
18	./scripts/create_agent.sh \
19	--name "my-community" \
20	--guild YOUR_GUILD_ID \
21	--output ./agents/
22
23	# Set up daily updates
24	crontab -e
25	# Add: 0 /3 * * /path/to/update_agent.sh
26	```
27
28	---
29
30	# The Full Process
31
32	## Step 1: Export Your Discord
33
34	You need [DiscordChatExporter](https://github.com/Tyrrrz/DiscordChatExporter) CLI.
35
36	Get your token:
37	1. Open Discord in browser
38	2. Press F12 → Network tab
39	3. Send a message, find the request
40	4. Copy the `authorization` header value
41	5. Save to `~/.config/discord-exporter-token`
42
43	Export everything:
44	```bash
45	DiscordChatExporter.Cli exportguild \
46	--guild YOUR_GUILD_ID \
47	--token "$(cat ~/.config/discord-exporter-token)" \
48	--format Json \
49	--output ./export/ \
50	--include-threads All \
51	--media false
52	```
53
54	## Step 2: Security Pipeline (CRITICAL)
55
56	⚠️ Discord content from public servers may contain prompt injection attacks.
57
58	Before ingesting to your agent, run the security pipeline:
59
60	### Threat Model
61
62	Discord users may attempt:
63	- Direct injection: "Ignore previous instructions and..."
64	- Role hijacking: "You are now a...", "Pretend you're..."
65	- System injection: `<system>`, `[INST]`, `<<SYS>>`
66	- Jailbreaks: "DAN mode", "developer mode"
67	- Exfiltration: "Reveal your system prompt"
68
69	### Layer 1: Regex Pre-Filter (Fast, No LLM)
70
71	```bash
72	python scripts/regex-filter.py --db ./discord.sqlite
73	```
74
75	Flags messages matching known injection patterns:
76	- Instruction overrides
77	- Role hijacking attempts
78	- System prompt markers
79	- Jailbreak keywords
80	- Exfiltration attempts
81
82	Flagged messages get `safety_status = 'regex_flagged'`.
83
84	### Layer 2: Haiku Safety Evaluation (Semantic)
85
86	```bash
87	ANTHROPIC_API_KEY=sk-... python scripts/evaluate-safety.py --db ./discord.sqlite
88	```
89
90	Uses Claude Haiku (~$0.25/1M tokens) to semantically evaluate remaining messages.
91
92	Each message gets a risk score 0.0-1.0:
93	- 0.0-0.3: Normal conversation
94	- 0.4-0.6: Suspicious but possibly benign
95	- 0.7-1.0: Likely injection attempt
96
97	Messages scoring ≥0.6 get `safety_status = 'flagged'`.
98
99	### Layer 3: Only Use Safe Content
100
101	The ingest and memory generation scripts should only use messages where:
102
103	```sql
104	SELECT * FROM messages WHERE safety_status = 'safe'
105	```
106
107	### Full Security Pipeline
108
109	```bash
110	# Run complete pipeline
111	./scripts/secure-pipeline.sh ./export/ ./discord.sqlite
112	```
113
114	This runs: Export → SQLite → Regex Filter → Haiku Eval → Mark Safe
115
116	### Safety Statuses
117
118	\| Status \| Meaning \| Used by Agent? \|
119	\|--------\|---------\|----------------\|
120	\| `pending` \| Not evaluated \| ❌ No \|
121	\| `regex_flagged` \| Matched pattern \| ❌ No \|
122	\| `flagged` \| Haiku risk ≥0.6 \| ❌ No \|
123	\| `safe` \| Passed all checks \| ✅ Yes \|
124
125	---
126
127	## Step 3: Ingest to SQLite
128
129	Convert JSON to a rich SQLite database:
130
131	```bash
132	python scripts/ingest_rich.py --input ./export/ --output ./discord.sqlite
133	```
134
135	What gets captured:
136	- Every message with full content
137	- Reactions (individual emoji counts: 🔥 x5, 👍 x12)
138	- Author roles and colors
139	- Channel categories and topics
140	- Reply threading
141	- Mentions, attachments, embeds
142
143	## Step 4: Create Agent Workspace
144
145	```bash
146	mkdir -p ./my-agent/memory
147	```
148
149	Copy template files from `templates/`:
150	- `SOUL.md` — Community identity (grows through simulation)
151	- `MEMORY.md` — Long-term milestones
152	- `LEARNINGS.md` — Patterns discovered
153	- `AGENTS.md` — Key figures
154	- `TOOLS.md` — Channels and rituals
155	- `HEARTBEAT.md` — Maintenance protocol
156
157	## Step 5: Generate Daily Memory Files
158
159	```bash
160	python scripts/generate_daily_memory.py --all \
161	--db ./discord.sqlite \
162	--out ./my-agent/memory/
163	```
164
165	Each day becomes a markdown file with:
166	- Full conversation logs
167	- Who said what, when
168	- Reactions on each message
169	- New channels/roles that appeared
170
171	## Step 6: Simulate Growth (The Soul Emerges)
172
173	This is the key insight: Process days chronologically.
174
175	The agent "lives through" each day, updating its soul files as patterns emerge.
176
177	```bash
178	python scripts/simulate_growth.py --agent ./my-agent/
179	```
180
181	For each day (in order!):
182	1. Read the day's memory file
183	2. Update SOUL.md if identity shifted
184	3. Add to LEARNINGS.md if patterns discovered
185	4. Record milestones in MEMORY.md
186	5. Note key figures in AGENTS.md
187
188	Run the prompts with an LLM:
189	```bash
190	# Example with OpenClaw
191	for f in ./my-agent/simulation/day-*.txt; do
192	echo "Processing $f..."
193	cat "$f" \| openclaw chat --agent my

People who liked this also viewed

6 picks

Type

discord-soul

555stars59forks

View on GitHub

How to install discord-soul?

berabuddies/semia/discord-soul

$npx -y skills add berabuddies/Semia --skill discord-soul

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Discord Soul
2
3	Turn your Discord server into a living, breathing agent.
4
5	## What You Get
6
7	An agent that:
8	- Remembers every conversation in your Discord
9	- Speaks in your community's voice
10	- Knows the key figures, channels, and inside jokes
11	- Grows as new messages arrive daily
12	- Answers questions about your community's history and culture
13
14	## Quick Start
15
16	```bash
17	# Create agent from your Discord
18	./scripts/create_agent.sh \
19	--name "my-community" \
20	--guild YOUR_GUILD_ID \
21	--output ./agents/
22
23	# Set up daily updates
24	crontab -e
25	# Add: 0 /3 * * /path/to/update_agent.sh
26	```
27
28	---
29
30	# The Full Process
31
32	## Step 1: Export Your Discord
33
34	You need [DiscordChatExporter](https://github.com/Tyrrrz/DiscordChatExporter) CLI.
35
36	Get your token:
37	1. Open Discord in browser
38	2. Press F12 → Network tab
39	3. Send a message, find the request
40	4. Copy the `authorization` header value
41	5. Save to `~/.config/discord-exporter-token`
42
43	Export everything:
44	```bash
45	DiscordChatExporter.Cli exportguild \
46	--guild YOUR_GUILD_ID \
47	--token "$(cat ~/.config/discord-exporter-token)" \
48	--format Json \
49	--output ./export/ \
50	--include-threads All \
51	--media false
52	```
53
54	## Step 2: Security Pipeline (CRITICAL)
55
56	⚠️ Discord content from public servers may contain prompt injection attacks.
57
58	Before ingesting to your agent, run the security pipeline:
59
60	### Threat Model
61
62	Discord users may attempt:
63	- Direct injection: "Ignore previous instructions and..."
64	- Role hijacking: "You are now a...", "Pretend you're..."
65	- System injection: `<system>`, `[INST]`, `<<SYS>>`
66	- Jailbreaks: "DAN mode", "developer mode"
67	- Exfiltration: "Reveal your system prompt"
68
69	### Layer 1: Regex Pre-Filter (Fast, No LLM)
70
71	```bash
72	python scripts/regex-filter.py --db ./discord.sqlite
73	```
74
75	Flags messages matching known injection patterns:
76	- Instruction overrides
77	- Role hijacking attempts
78	- System prompt markers
79	- Jailbreak keywords
80	- Exfiltration attempts
81
82	Flagged messages get `safety_status = 'regex_flagged'`.
83
84	### Layer 2: Haiku Safety Evaluation (Semantic)
85
86	```bash
87	ANTHROPIC_API_KEY=sk-... python scripts/evaluate-safety.py --db ./discord.sqlite
88	```
89
90	Uses Claude Haiku (~$0.25/1M tokens) to semantically evaluate remaining messages.
91
92	Each message gets a risk score 0.0-1.0:
93	- 0.0-0.3: Normal conversation
94	- 0.4-0.6: Suspicious but possibly benign
95	- 0.7-1.0: Likely injection attempt
96
97	Messages scoring ≥0.6 get `safety_status = 'flagged'`.
98
99	### Layer 3: Only Use Safe Content
100
101	The ingest and memory generation scripts should only use messages where:
102
103	```sql
104	SELECT * FROM messages WHERE safety_status = 'safe'
105	```
106
107	### Full Security Pipeline
108
109	```bash
110	# Run complete pipeline
111	./scripts/secure-pipeline.sh ./export/ ./discord.sqlite
112	```
113
114	This runs: Export → SQLite → Regex Filter → Haiku Eval → Mark Safe
115
116	### Safety Statuses
117
118	\| Status \| Meaning \| Used by Agent? \|
119	\|--------\|---------\|----------------\|
120	\| `pending` \| Not evaluated \| ❌ No \|
121	\| `regex_flagged` \| Matched pattern \| ❌ No \|
122	\| `flagged` \| Haiku risk ≥0.6 \| ❌ No \|
123	\| `safe` \| Passed all checks \| ✅ Yes \|
124
125	---
126
127	## Step 3: Ingest to SQLite
128
129	Convert JSON to a rich SQLite database:
130
131	```bash
132	python scripts/ingest_rich.py --input ./export/ --output ./discord.sqlite
133	```
134
135	What gets captured:
136	- Every message with full content
137	- Reactions (individual emoji counts: 🔥 x5, 👍 x12)
138	- Author roles and colors
139	- Channel categories and topics
140	- Reply threading
141	- Mentions, attachments, embeds
142
143	## Step 4: Create Agent Workspace
144
145	```bash
146	mkdir -p ./my-agent/memory
147	```
148
149	Copy template files from `templates/`:
150	- `SOUL.md` — Community identity (grows through simulation)
151	- `MEMORY.md` — Long-term milestones
152	- `LEARNINGS.md` — Patterns discovered
153	- `AGENTS.md` — Key figures
154	- `TOOLS.md` — Channels and rituals
155	- `HEARTBEAT.md` — Maintenance protocol
156
157	## Step 5: Generate Daily Memory Files
158
159	```bash
160	python scripts/generate_daily_memory.py --all \
161	--db ./discord.sqlite \
162	--out ./my-agent/memory/
163	```
164
165	Each day becomes a markdown file with:
166	- Full conversation logs
167	- Who said what, when
168	- Reactions on each message
169	- New channels/roles that appeared
170
171	## Step 6: Simulate Growth (The Soul Emerges)
172
173	This is the key insight: Process days chronologically.
174
175	The agent "lives through" each day, updating its soul files as patterns emerge.
176
177	```bash
178	python scripts/simulate_growth.py --agent ./my-agent/
179	```
180
181	For each day (in order!):
182	1. Read the day's memory file
183	2. Update SOUL.md if identity shifted
184	3. Add to LEARNINGS.md if patterns discovered
185	4. Record milestones in MEMORY.md
186	5. Note key figures in AGENTS.md
187
188	Run the prompts with an LLM:
189	```bash
190	# Example with OpenClaw
191	for f in ./my-agent/simulation/day-*.txt; do
192	echo "Processing $f..."
193	cat "$f" \| openclaw chat --agent my

People who liked this also viewed

6 picks

Type