agentleFS
nuffinnuffin/hermes-task-framework/skills/task-context-storage/SKILL.md

task-context-storage

Skill2 starsChanged 45 days agoMITView on GitHub

Route facts among global memory, task MEMORY.md, and task CHANGELOG.md.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    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:

    Do not store task progress, commits, issue numbers, transient failures, or session outcomes here.

    line 51
  2. Distinctive:

    Do not store momentary runtime status or chronological history here.

    line 59
  3. Very specific:

    When a task spans multiple subsystems, load compact-directory-memory and use its hierarchical layout.

    line 71
  4. Distinctive:

    When one observation contains both a reusable preference and a task decision, split it:

    line 80

≈ “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

Task Context Storage

Defines which persistence layer owns each kind of context. File formats, nested directory structure, and maintenance tooling are delegated to compact-directory-memory.

Persistence layers

LayerScopeContentRead/write timing
Global memory()All sessions and tasksUser preferences and stable environment factsAutomatically injected; update sparingly
Task MEMORY.mdOne task or subsystemStable expected-state facts and constraintsRead fully before work; update only when facts change
Task CHANGELOG.mdOne task or subsystemDecisions, operations, observed verification, blockers, next stepRead recent entries before work; append after work
Raw logsOne executionCommand output, traces, generated diagnosticsRead as needed; safe to clean according to task rules

Route to global memory()

  • User identity, durable preferences, and recurring corrections.
  • Stable environment facts used across unrelated tasks.
  • Cross-task workflow constraints.

Do not store task progress, commits, issue numbers, transient failures, or session outcomes here.

Route to task MEMORY.md

  • Expected configuration and stable conventions.
  • Durable paths, interfaces, ports, ownership boundaries, and invariants.
  • Facts that another session must know before changing the same task or subsystem.

Do not store momentary runtime status or chronological history here.

Route to task CHANGELOG.md

  • Why one option was chosen over another.
  • What was changed and where.
  • Real test, build, migration, or health-check results.
  • Failures, fixes, unresolved blockers, and exact next step.
  • Session and phase outcomes.

Hierarchical tasks

When a task spans multiple subsystems, load compact-directory-memory and use its hierarchical layout. Root files index and synthesize cross-subsystem facts; memories/<sub-system>/ owns detailed subsystem context.

The routing rule remains unchanged at every level:

  • stable expected state → MEMORY.md
  • chronological observed work → CHANGELOG.md

Boundary cases

When one observation contains both a reusable preference and a task decision, split it:

  • global memory() receives the durable preference;
  • task CHANGELOG.md receives the local decision and its consequences;
  • task MEMORY.md receives only a resulting stable fact, if one exists.

Relationship to task-framework

task-framework should reference this skill for context-routing decisions and compact-directory-memory for concrete directory management. It should not duplicate either skill's detailed rules.

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.

Reviews can't be read right now.

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.