agentleFS
matthewdonsemail-labmatthewdonsemail-lab/log/skills/runs/SKILL.md

runs

Skill63 starsChanged 2 days agoMITView on GitHub

Keep the recorded runs from filling the disk (rule 66). `log runs` lists every run under ~/.logCli/runs with its size, whether rea settled it and whether the request docs hold it; `log runs prune --apply` deletes the analysed ones, keeping each one's findings, steps and requests by domain. Use when the storage hook speaks, AdsPower says it is running out of disk space, or the owner asks to clear the runs.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    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:

    Every action records a raw HAR (rule 19): a warm-up run is about 1.5 GB, an inspect run up to 1.4 GB.

    line 8
  2. Very specific:

    Not in the docs: log capture docs, then write any request's reasoning it asks for.

    line 18

≈ “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

runs

Every action records a raw HAR (rule 19): a warm-up run is about 1.5 GB, an inspect run up to 1.4 GB. On 2026-10-22 they reached 12.5 GB, the disk had 2 GB left and AdsPower refused to start Bethlehem's browser.

The loop

  1. Analyse (rule 25): a run may go only when rea settled it (settle.json: every step read, findings.md written) and the request docs hold its requests (docs/platforms/facebook/operations.json built after it).
    • Not settled: log capture settle <its capture folder>.
    • Not in the docs: log capture docs, then write any request's reasoning it asks for.
  2. Look: log runs lists each run with may go or why it is kept (still being written, not settled, not in the docs).
  3. Prune: log runs prune --apply (--before <YYYY-MM-DD> keeps the runs from that day on). First it takes every step's internal API calls out of its HAR (exchanges.json: what was sent and what came back, rule 73), the same as log runs exchanges does for every run on disk. A run with a step HAR it can't read is not pruned. Each pruned run leaves:

    • ~/.logCli/runs/kept/<action>/<run>/: findings.md, steps.jsonl, settle.json, pruned.json, goalFindings.md and .jsonl, every inspect step's steps/<step>/exchanges.json, reading.json and readingState.json, and every settle cut's <browser>/steps/<step>.exchanges.json (2 MB for 11 steps, against 1.2 GB of HAR);
    • one line in ~/.logCli/runs/pruned.jsonl: size, settled, documented, and its requests counted by docs domain (groups, reels, checkpoint…).

    A request's doc keeps what its pruned runs showed. The docs build folds every request's fields and sightings into ~/.logCli/rea/clientFacts.json (local: run ids carry account ids). Each run is counted once, a request seen again only adds its new rows, and a field a later build couldn't read (the bundle wasn't captured) keeps what was known. Without the store, a request starts from its committed doc (factsFromDoc). The store also keeps each request's newest 3 real exchanges (one per step: what it sent and what came back, secrets and signed link tokens out, each cut to 6,000 characters by depth), and its doc shows them under "What it sent and what came back". First seen 2026-10-22: a rebuild after the prune had cut FBYRPTimeLimitsEnforcementQuery from 17 runs to 1.

The hook

scripts/hooks/storage.ts (SessionStart, UserPromptSubmit, src/agentHooks/registry) says in the agent's context when ~/.logCli/runs holds over 2 GB or the disk has under 5 GB free (storageVerdict), measured at most every 15 minutes. log runs space prints the same.

Tests

bun test loads scripts/tests/preload.ts (bunfig.toml), which points LOG_RUNS_DIR at a scratch folder: tests never write into the real runs again (the profile tests had left about 76 galwayOne folders there).

Code: src/browser/runStore, src/cli/runs. The requirements: docs/specs/runsAndRea.md.

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.