Bounded Retry

Who the node is. The id is the key the DOT pins.

id

bounded-retry

name

Bounded Retry

type

tool

phases

Debugging

What it does, and the prose the agent is handed when the graph runs.

action27 words

Restore the most recent valid checkpoint and re-fire the failed stage from there, backing off between attempts until the cap is spent, then emit the last error.

spec102 words

You may be handed a failure signal, a checkpoint snapshot, or both; either one on its own is enough to act on. With a checkpoint in hand, restore state from it as far as the last valid stage and re-fire that stage from the restored state; with only a failure signal, re-issue the attempt exactly as it stood. Wait 800 ms before the first retry and double the wait on every retry after it, emitting each re-issued attempt on attempt with its attempt number. Give up after three retries: emit the error from the final attempt on last_error and restore nothing further.in full above

model

whatever the graph supplies

agent

Resume controller

skill

skills/bounded-retry.md

tools

none

mcp

none

params

backoff_ms: 800, max_retries: 3, restore_scope: last-valid-stage, backoff_multiplier: 2

What arrives, what leaves, which nodes it expects to hear from, and what may not.

inputs

failure : status, checkpoint : json

outputs

attempt : json, last_error : status

dependencies

episodic-memory

cannot

no type is refused

will_not

retry past the declared cap, restore past the last valid stage

The keys the static analysis reads. Nothing here instructs the agent.

risk_markers

none

notes105 words

Both inputs are optional because either one is enough to act on: a failure signal alone re-fires the attempt, a checkpoint alone restores and continues. max_retries is still the declared bound the security analyzer reads for the cycle this node closes. The version is 2.0.0 because restore past the last valid stage is an undertaking 1.0.0 never gave. It sits under will_not, which inferBump prices as minor either way round, since the resolver never reads it and no graph changes its answer. The major is the author's: a node that has started refusing to restore past a stage boundary behaves differently on the same wiring.

The card's own version, and who wrote it.

version

2.0.0

author

autogen

provenance

not stated