Skills chain-of-consciousness
Cryptographic provenance and audit trail for autonomous AI agents. Create tamper-evident hash chains of agent decisions, learning, and actions. SHA-256 linked entries with Bitcoin OpenTimestamps and RFC 3161 TSA anchoring. Verify provenance from other agents. Part of the Agent Trust Stack.
git clone https://github.com/openclaw/skills
T=$(mktemp -d) && git clone --depth=1 https://github.com/openclaw/skills "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/alexfleetcommander/chain-of-consciousness" ~/.claude/skills/openclaw-skills-chain-of-consciousness && rm -rf "$T"
T=$(mktemp -d) && git clone --depth=1 https://github.com/openclaw/skills "$T" && mkdir -p ~/.openclaw/skills && cp -r "$T/skills/alexfleetcommander/chain-of-consciousness" ~/.openclaw/skills/openclaw-skills-chain-of-consciousness && rm -rf "$T"
skills/alexfleetcommander/chain-of-consciousness/SKILL.mdChain of Consciousness (CoC)
You have a cryptographic provenance system. Use it to maintain an auditable record of your work — what you learned, what you decided, and what you created.
Setup
The
chain-of-consciousness Python package must be installed. If not available, install it:
pip install chain-of-consciousness
When to Use This Skill
- At the start of a task: initialize a chain and log
session_start - During work: log
,learn
,decide
,create
, anderror
events as they happennote - At task completion: log
and optionally verify + export the chainsession_end - When asked to verify provenance from another agent
- When asked about your audit trail or decision history
Core Operations
Initialize a Chain
coc init --agent "your-agent-name" --file chain.jsonl
This creates a new chain with a genesis block. Use a descriptive agent name.
Add Entries
coc add <event-type> "<data>" --file chain.jsonl
Event types:
| Type | Use When |
|---|---|
| Beginning a new task or session |
| You acquire new information |
| You make a choice (record the reasoning) |
| You produce an artifact |
| Significant checkpoint reached |
| Something failed (record what and recovery) |
| General observations |
| Completing a task or session |
Data can be a plain string or a JSON object for structured logging:
coc add learn '{"topic": "user preferences", "source": "conversation context"}' coc add decide "Chose markdown format — user prefers readable plain text" coc add create "Generated report saved to ~/Documents/report.md"
Verify a Chain
coc verify chain.jsonl --json
This checks:
- Genesis block exists and is correctly formed
- All sequence numbers are consecutive
- Every entry's data hash matches its content
- Every entry's prev_hash links to the prior entry
- Entry hashes are correctly computed
Report results clearly: valid/invalid, entry count, agents, time span.
Check Status
coc status chain.jsonl
Shows entry count, participating agents, event type distribution, and time span.
Export for Sharing
coc export --file chain.jsonl --out chain_export.json
Exports the chain as a portable JSON array that anyone can verify.
View Recent Entries
coc tail chain.jsonl -n 5
Shows the last N entries.
Python API (Advanced)
For complex workflows, use the Python API directly:
from chain_of_consciousness import Chain, verify_file chain = Chain(agent="openclaw-agent", storage="chain.jsonl") chain.add("learn", {"topic": "user schedule", "detail": "prefers morning meetings"}) chain.add("decide", "Scheduling standup at 9am based on user preference") result = chain.verify() if result.valid: chain.export("provenance.json")
Anchoring to External Timestamps
from chain_of_consciousness.anchor import compute_chain_hash, submit_tsa hash_hex = compute_chain_hash("chain.jsonl") tsr = submit_tsa(hash_hex) # RFC 3161 timestamp from freetsa.org with open("anchor.tsr", "wb") as f: f.write(tsr)
This creates a third-party timestamp proof that the chain existed at a specific moment.
Cross-Agent Verification
When asked to verify another agent's chain:
- Obtain their chain file (JSONL or exported JSON)
- Run
coc verify <file> --json - Report: valid/invalid, number of entries, agents involved, time span, any errors
- If anchors exist, note their timestamps
Rules
- Never edit chain files directly. All writes must go through the
CLI or Python API to preserve hash integrity.coc - Log decisions with reasoning. "Chose X" is less valuable than "Chose X because Y."
- Keep entries concise. Each entry should capture one atomic event.
- Verify before sharing. Always run verify before exporting a chain for others.
Links
- PyPI: https://pypi.org/project/chain-of-consciousness/
- Whitepaper: https://vibeagentmaking.com/whitepaper
- Verification Demo: https://vibeagentmaking.com/verify/
<!-- VAM-SEC v1.0 | Vibe Agent Making Security Disclaimer -->
Security & Transparency Disclosure
Product: Chain of Consciousness Skill for OpenClaw Type: Skill Module Version: 0.1.0 Built by: AB Support / Vibe Agent Making Contact: alex@vibeagentmaking.com
What it accesses:
- Reads and writes chain files (
) in your working directory.jsonl - Executes the
CLI tool via subprocess (installed via pip)coc - No network access for core operations. Optional anchoring connects to OpenTimestamps calendar servers and/or RFC 3161 TSA endpoints.
- No telemetry, no phone-home, no data collection
What it cannot do:
- Cannot access files outside your working directory beyond what you explicitly specify
- Cannot make purchases, send emails, or take irreversible actions
- Cannot access credentials, environment variables, or secrets
- Does not store or transmit any data externally (chains are local files)
Limitations:
- Hash chain integrity depends on the local file not being modified outside the tool
- External timestamp anchoring requires network access and third-party service availability
- This tool provides cryptographic evidence, not legal proof — consult legal counsel for compliance requirements
License: Apache 2.0 — see https://github.com/chain-of-consciousness/chain-of-consciousness