agentleFS
Sign inStart free
dynos-fitdynos-fit/dynos-work/skills/trajectory/SKILL.md

trajectory

Skill2 starsChanged 8 months agoMITView on GitHub

Internal dynos-work skill. Sequence memory manager. Stores compact task traces and retrieves similar prior tasks to inform discovery and design review. Runs only when the user explicitly types /dynos-work:trajectory; never auto-triggered from conversation.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 8 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If current code contradicts memory, trust the repo and call the memory mismatch out.

    line 13
  2. Distinctive:

    Do not store or retrieve traces so loosely that they become decorative noise.

    line 14
  3. Distinctive:

    If available in this repo, the deterministic runtime for this skill is:

    line 16
  4. Very specific:

    State Signature Search: Call the state-encoder to get the current module's state signature ($).

    line 33
  5. Very specific:

    Trajectory Search: Use the deterministic similarity search in hooks/trajectory.py to find the 3 most similar successful trajectories in …

    line 34
  6. Distinctive:

    When a similar prior task is found:

    line 37
  7. Distinctive:

    Use it to surface likely failure modes, test gaps, and architecture risks.

    line 38
  8. Distinctive:

    Never copy its implementation plan or decision path blindly.

    line 39
  9. Distinctive:

    If current repo evidence conflicts with retrieved history, trust current repo evidence.

    line 40
  10. Distinctive:

    Trajectory Integrity: Never modify a past trajectory.

    line 43

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

dynos-work: Trajectory (Sequence Memory)

This is the memory layer for prior task traces. It is retrieval support, not an autonomous policy engine.

Ruthlessness Standard

  • Retrieved history is evidence of recurrence, not permission to cargo-cult old solutions.
  • If current code contradicts memory, trust the repo and call the memory mismatch out.
  • Do not store or retrieve traces so loosely that they become decorative noise.

If available in this repo, the deterministic runtime for this skill is:

python3 hooks/trajectory.py rebuild --root .
python3 hooks/trajectory.py search query.json --root . --limit 3

What you do

Step 1 -- Task Trace Extraction (PUSH)

Immediately after a successful task completion and /dynos-work:learn:

  1. Trace Reconstruction: Reconstruct the sequence of (State, Action, Reward) for the entire task.
  2. Update Registry: Push the sequence into the versioned store at .dynos/trajectories.json.
  3. Outcome Summary: Record enough metadata to help future retrieval find similar successful tasks.

Step 2 -- Behavioral Retrieval (PULL)

During the "Founder Phase" or "Planning Phase":

  1. State Signature Search: Call the state-encoder to get the current module's state signature ($).
  2. Trajectory Search: Use the deterministic similarity search in hooks/trajectory.py to find the 3 most similar successful trajectories in trajectories.json.

Step 3 -- Advisory Use

When a similar prior task is found:

  1. Use it to surface likely failure modes, test gaps, and architecture risks.
  2. Never copy its implementation plan or decision path blindly.
  3. If current repo evidence conflicts with retrieved history, trust current repo evidence.

Hard Rules

  • Trajectory Integrity: Never modify a past trajectory; they are immutable ground truth.
  • Anonymization: Strip all local paths and IDs from the trajectories before they are indexed for search.
  • Advisory Only: Retrieval may influence prioritization, but it never overrides hard validation or human approval.
  • Schema First: Trajectories must be written in the versioned store format. Do not append ad hoc JSON blobs.

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.

Check your setup, free