agentleFS
VorElzeVorElze/hermes-office-skills/skills/hermes-stream-stale-diagnosis/SKILL.md

hermes-stream-stale-diagnosis

Skill0 starsChanged 21 days agoMITView on GitHub

Use when Hermes reports stale provider streams or timeouts.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use CLI only; never edit YAML by hand:

    line 34
  2. Very specific:

    Verify effective values by calling the real timeout resolver with the same HERMES_HOME and verify compression via config loader.

    line 45
  3. Distinctive:

    Do not configure fallback until the alternate provider credential and suitable model are verified.

    line 56

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Diagnose Hermes stream-stale failures

When to Use

Load this when Hermes reports Stream stale for Ns, No response from provider after Ns, ReadError('Error reading post'), or Could not open a stream. Skip it for Telegram ConnectTimeout (network path, not provider stream) — check gateway transport logs instead.

Proven diagnosis pattern

  • Inspect per-profile request dumps (sessions/request_dump_*.json), not just gateway logs. Dumps can identify independent rate_limit_error / HTTP 429 failures.
  • A measured Koordinator request was 1,709,991 bytes with 1,121 messages. Replaying it against the configured endpoint returned SSE message_start after 37.7s. An explicit 120s timeout can override Hermes' context-aware default for large prompts; resolve the effective timeout with hermes_cli.timeouts.get_provider_stale_timeout, not by reading YAML alone.
  • Distinguish 429 short-window limits from a slow prefill; they require separate mitigations. Hermes classifier already honors retry_after and has fallback classification.

Safe prevention changes

Use CLI only; never edit YAML by hand:

HERMES_HOME=~/.hermes/profiles/<profile> hermes config set providers.anthropic.request_timeout_seconds 300
HERMES_HOME=~/.hermes/profiles/<profile> hermes config set providers.anthropic.stale_timeout_seconds 300
HERMES_HOME=~/.hermes/profiles/<profile> hermes config set compression.threshold_tokens 60000

Repeat for each intended profile. For global settings set HERMES_HOME=~/.hermes explicitly; the active session's default HERMES_HOME may be a profile home.

Verify effective values by calling the real timeout resolver with the same HERMES_HOME and verify compression via config loader. The multiplexer watches profile config.yaml/.env signatures and reconciles changed profiles; confirm Re-scanned profile in gateway logs before considering a restart. Do not restart a live gateway without need/authorization.

Avoid unsafe fallback assumptions

fallback_providers: [] means no fallback. Adding one through hermes fallback add uses an interactive provider/model picker (_require_tty) and needs a valid credential for a distinct provider. Never assume an env key exists from an old session or config comment; check only key names and never print values. A fallback using the same shared rate-limited API key may not relieve 429s. Do not configure fallback until the alternate provider credential and suitable model are verified.

Pitfalls

  • Correct Anthropic-compatible route in this deployment: /anthropic/v1/messages.
  • Old request dumps can contain expired Authorization tokens; use a currently valid secret from the appropriate secret store, without logging it.
  • Telegram ConnectTimeout is a separate network-path issue; check gateway recovery logs, don't confuse it with model provider timeout or API rate-limit errors.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.