agentleFS
jhutchison0jhutchison0/tacsop/.claude/skills/maintaining-project-context/SKILL.md

maintaining-project-context

Skill0 starsChanged 25 days agoApache-2.0View on GitHub

Maintain CONTEXT.md at the repo root capturing the project's identity, mission, current state, and key constraints. Use at the start of significant new work, when the project mission or scope changes, when an agent needs to understand the project quickly, or when the project's relationship with downstream consumers changes.

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 invoke this skill for:

    line 23
  2. Distinctive:

    Do not duplicate content across these files.

    line 75
  3. Distinctive:

    When CONTEXT.md needs an update:

    line 81
  4. Very specific:

    Cross-check config/project.yaml: If your update touches identity (version, phase), make sure the YAML still agrees.

    line 86

≈ “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

Maintaining Project Context

CONTEXT.md is a one-page narrative of what a project is, what it cares about, and the constraints on how it operates. It is the file you would hand to a new contributor (human or agent) who has 60 seconds to understand the project well enough to act in alignment with it.

Adapted from Matt Pocock's grill-with-docs pattern, distinct from LANGUAGE.md (which defines terms) and config/project.yaml (which holds machine-readable identity).

When to Use

Invoke this skill when:

  • Significant new work begins: a new phase, a new mission, a structural change to how the project relates to its consumers.
  • The project's mission or scope materially shifts (not for routine task progress).
  • A new agent or human contributor needs to ramp up quickly and needs project orientation.
  • The relationship with downstream consumers changes (new repos joining, repos leaving, propagation protocol changes).
  • A key constraint is discovered or relaxed (e.g., a hard rule becomes a soft preference, or vice versa).

Do not invoke this skill for:

  • Routine task completion or session progress — those go in session docs.
  • Code-level decisions — those are commits or, when warranted, ADRs.
  • Status updates — those are SITREP / status-report.

CONTEXT.md Format

The canonical sections, in order:

# CONTEXT.md — What This Project Is

[One-paragraph orientation. What this project is. Why it exists.]

## Identity
[1–2 paragraphs. What the project IS, not what it does. Distinguishing facts.]

## Mission
[Bullet list. Goals at the project level, not task level.]

## Current State
[Where each part of the current state lives, never a snapshot of it: version and phase in config/project.yaml, the campaign focus in docs/tasks.md, known unknowns in docs/gaps.md, the last session in docs/sessions/, and the command that measures anything that drifts. A count or a "most recent" written here reads as current after it stops being true.]

## Constraints
[Pillars, hard rules, anti-rules. Things that govern decision-making.]

## Key Relationships
[Downstream consumers, agent roster summary, team templates.]

## Reading Order for New Agents and Contributors
[Numbered list of files to read, in order.]

## Distinguishing This File from Adjacent Artifacts
[Small table: CONTEXT vs LANGUAGE vs project.yaml vs CLAUDE.md vs sessions vs plans.]

The body should fit on one or two screens. If CONTEXT.md is creeping past ~250 lines, it's drifting into spec territory: extract the detail to a design doc and link.

Distinguishing CONTEXT.md from Adjacent Artifacts

ArtifactWhat it answers
CONTEXT.mdWhat is this project and what does it care about?
LANGUAGE.mdWhat do specific terms mean in this project?
config/project.yamlWhat is the project's machine-readable identity?
docs/tasks.mdWhat comes next, and in what order? (the ## Focus section)
docs/gaps.mdWhat can nobody see yet, and what does that block?
CLAUDE.mdWhat conventions and commands does the agent need at hand?
docs/sessions/What was just worked on?
docs/plans/What is planned next?
docs/adr/Why was a hard-to-reverse decision made?

Do not duplicate content across these files. When tempted, ask: "Which one of these is this really about?" and put it there.

CONTEXT.md is the narrative counterpart to config/project.yaml's structured identity. They reinforce each other. If they disagree on version or phase, treat project.yaml as authoritative and update CONTEXT.md. Neither holds measured state: that is measured by command when it is needed (the maintaining-the-common-operating-picture skill).

Workflow

When CONTEXT.md needs an update:

  1. Identify the trigger. Which one of the "When to Use" conditions fired?
  2. Locate the right section. Identity, Mission, Current State, Constraints, Key Relationships, or Reading Order? Most updates touch one section.
  3. Make the minimum change. A new constraint = one new bullet. A new downstream relationship = one row in the Key Relationships section.
  4. Cross-check config/project.yaml. If your update touches identity (version, phase), make sure the YAML still agrees. If not, update the YAML in the same change. Active work is not here: it is the ## Focus section of docs/tasks.md.
  5. Update the Last Updated date at the bottom of CONTEXT.md.
  6. Confirm with the user. Changes to CONTEXT.md are policy-level. Always confirm before committing.

What CONTEXT.md Is Not

  • A status report (use SITREP / status-report).
  • A specification of any specific feature (use design docs in docs/design/).
  • A glossary of terms (use LANGUAGE.md).
  • A history of what changed (use docs/sessions/ and git log).
  • A list of every constraint imaginable (only the ones that actively govern decisions belong).

Sources

  • Matt Pocock, mattpocock/skills/skills/engineering/grill-with-docs/ — the inspiration; we adapted format and discipline.
  • This project's CONTEXT.md — the artifact this skill maintains.
  • This project's LANGUAGE.md — companion glossary.
  • This project's config/project.yaml — machine-readable identity.

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.