Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/sandeep-alluru/agentdelta/recordgit clone --depth 1 https://github.com/sandeep-alluru/agentdeltaWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00000 | $0.00415 |
| Opus 5 | $0.00000 | $0.00208 |
| Sonnet 5 | $0.00000 | $0.00083 |
| Haiku 4.5 | $0.00000 | $0.00042 |
Grade A, and why
record scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Generate boilerplate code to record an agent run with agentdelta.
Usage: /project:record [framework] [agent_variable_name]
Arguments: $ARGUMENTS (optional: framework name like "langchain", "langgraph", "custom")
Generate a self-contained Python snippet that:
For LangChain/LangGraph (default):
from agentdelta import record
# Baseline (before your change)
with record("baseline.jsonl", run_id="v1.0") as cb:
agent.invoke({"input": "..."}, config={"callbacks": [cb]})
# Candidate (after your change)
with record("candidate.jsonl", run_id="v1.1") as cb:
agent.invoke({"input": "..."}, config={"callbacks": [cb]})
For custom/framework-agnostic:
from agentdelta import AgentTrace
from agentdelta.trace import TraceNode, TraceEdge, NodeType, EdgeType
trace = AgentTrace(run_id="my_run")
trace.add_node(TraceNode(step=1, node_type=NodeType.START, content="user input here"))
trace.add_node(TraceNode(step=2, node_type=NodeType.LLM, content="reasoning text here"))
trace.add_node(TraceNode(step=3, node_type=NodeType.TOOL_CALL, content="tool_name(args)"))
trace.add_node(TraceNode(step=4, node_type=NodeType.TOOL_RETURN, content="tool result"))
trace.add_node(TraceNode(step=5, node_type=NodeType.END, content="final output"))
trace.save("my_run.jsonl")
Then show the diff command: agentdelta diff baseline.jsonl candidate.jsonl
Context:
- AgentdeltaCallback is LangChain BaseCallbackHandler-compatible (no import of BaseCallbackHandler needed)
- record() saves the trace on context exit, even if the agent raises an exception
- run_id appears in the diff report header — use something meaningful (version, git hash, etc.)
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 42 lines · 0 tokens per session scan A 2ee7ea1877a1
record is a command published in the GitHub repository sandeep-alluru/agentdelta (0 stars, last pushed 16d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 415 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
pre-commit
Run pre-commit checks identical to GitHub CI (ruff, mypy, tests).
off
Turn diffscope's automatic change briefing off for this repository (or everywhere with --global).
pensyve
Route explicit Pensyve memory requests through the bundled MCP server; supports recall, remember, observe, inspect, status, review, forget, and mention-style guidance.
inspect
View all memories stored for an entity in Pensyve.
remember
Store a fact about an entity in Pensyve memory.
date
询问当天的日期,输出的格式为 yyyy-MM-dd 星期几.